QPI assay of fibroblasts resilience to adverse effects of nanoGO clusters by multimodal and multiscale microscopy

Graphene is considered a possible drug deliver in nanomedicine for its mechanical, physical and chemical characteristics. Thus, studying graphene biocompatibility is pivotal to contribute to the modern nano-therapy science. The coexistence between cells and graphene should be analysed using non-invasive technologies and thus quantitative phase imaging (QPI) modalities are suitable to investigate the morphometric evolution of cells under nanomaterial exposure. Here, we show how a multimodal QPI approach can furnish a noninvasive analysis for probing the dose-dependent effect of nanoGO clusters on adherent NIH 3T3 fibroblast cells. We rely on both digital holography and Fourier ptychography (FP) in transmission microscopy mode. The former allows accurate time-lapse experiments at the single cell level. The latter provides a wide field of view characterization at the cells network level, thus assuring a significant statistical measurement by exploiting the intrinsic large space-bandwidth product of FP. The combination of these two techniques allows one to extract multimodal information about the cell resilience to adverse effects of nanoGO in the surrounding buffer, namely through quantitative, multi-scale, and time-resolved characterization.


Introduction
Nowadays graphene is the carbon allotrope that is attracting the greatest interest for its wide variety of potential applications.In fact, the 2D structure of carbon atoms arranged in a hexagonal lattice allows one to achieve unique physical and chemical properties [1][2][3].The GO is well known as a graphene species made of a planar carbon structure with oxygen functional groups at high density.The unique physical/chemical properties of GO have attracted great interest in the last years and different applications have been proposed in the literature [1,4,5].In particular, the nano-GO (nGO) is obtained from GO by converting the micrometric lateral dimensions of the GO sheets to nanometric size [6] and the ratio between the elements C/O determines the level of oxidation [7,8].The oxygen functional groups allow one to perform various functionalization approaches [7,9], thanks to the negative and positive reactive sites achievable [10].In fact, the π-conjugated structure of GO can be used for capturing drug molecules through π-π stacking interactions or by efficient covalent bonding [7], thus making it an interesting choice as a nanocarrier for biochemical molecules [11][12][13][14].nGO is nowadays considered as the next generation of carbon materials that will be used as anticancer therapeutic agent [15][16][17][18][19][20].
In this framework, the interaction of nGO carriers with in-vitro cell population must be carefully studied especially in terms of induced toxicity to translate on patients a well-established technology for therapy [14].To this aim, the knowledge of the nGO effect in relation to their concentration and degree of oxidation is highly demanding.Nanotoxicology and nanomedicine make extensive use of in-vitro cellular assays, such as flow cytometry-based assay using fluorophores [21], Comet assay [22,23], Alamar Blue, neutral red (NR), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), and 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2 H-tetrazolium (WST-1) [24][25][26], which were developed prior to the nanotechnology era and the introduction of graphene-based nanomaterials affects the accuracy of this type of standard assays, due to the interference of nano-particles with the above-mentioned assays.In general, NPs can interfere with optical measurements through light absorption interacting with dyes used as molecular probes of cellular integrity [27,28].In some cases, the resulting artifacts can lead to a significant misinterpretation of effects on cell viability and cytotoxicity [29].This means that cell toxicity profile evaluation is strictly dependent on the method, the protocol and the cell type used.
The optical analysis at single cell level is usually carried out through different methods, based on both single shot and time-lapse observation.Within the wide plethora of optical microscopy techniques, the stain-based methods rely on the detection of the fluorescence signal emitted by specific regions in the cells or by intracellular organelles.However, the fluorophores used in MTT [17,28,30] assay for the cell viability can interfere with the graphene-based NPs [31,32].Moreover, the accurate specificity of fluorescence-based probing is paid in terms of invasiveness, phototoxicity and photobleaching.These effects might prevent in some cases a reliable long time-lapse observations.On the contrary, in label-free optical microscopy approaches based on Quantitative phase imaging (QPI) [33][34][35] the samples are imaged without stains or dyes [36].Therefore, QPI techniques are good candidates for rapid, noninvasive, and quantitative investigation of the morphometric changes induced in a cell population when exposed to graphene-based nanomaterial, thus avoiding the interference with standard in-vitro cell assay.QPI microscopes measure the optical path delay that the sample introduces on the transmitted wavefront, which conveys quantitative information related to biophysical parameters of the specimen.Optical thickness, refractive index distributions, biovolume and dry mass are physical quantities measurable from the optical readout [37,38].Nevertheless, the inspection of the interplay between cells and nGO particles requires multi-modal approaches to address the trade-off between lateral resolution, time resolution, depth of focus and FoV.Indeed, large lateral resolution is required to detect the presence of nGO particles and to correlate their positions to the cells centroids.On the other hand, large FoV is needed to gather an ensemble view of the cells layout, to link it to the nGO centroids, and to obtain statistically relevant information.In this framework, two label-free QPI techniques outstand, namely digital holography (DH) in microscopy [32,39] and FP [40][41][42][43].
Here, we show how DH and FP can be fruitfully combined to study quantitatively the effect of nGO clusters on adherent fibroblast cells especially in terms of life and death process.In our previous publication [31], we focused the attention on the viability of cells after exposure to nGO with three different oxidation degrees (named nGO1, nGO2, nGO3) and on nGO internalization.The study was made using MTT, phase contrast under a benchtop microscope, and DH microscopy.nGO3 provoked the lowest viability level according to the MTT test, and only nGO2 was imaged under the DH microscope.Here, we investigate the cells resilience to nGO3 clusters, i.e. the highest oxidation degree investigated in [31], that are not necessarily internalized.At this scope, we implement a new analysis which combines DH in time lapse mode and the novel FP method.It is important to notice that, unlike the previous work, we perform a comprehensive study of label-free DH parameters in time.We show the complementarity between the two microscopes for assessing the viability of cells under label-free mode.Moreover, we concern the cytotoxicity profile of the nGO3, taking into account the high density of reactive species and hence the ability to carry a high amount of functional groups for drug delivery purposes.
Two different doses, 20 µg ml −1 and 100 µg ml −1 were added to the live cell medium for studying the corresponding effects on live cells.Here, DH allows us to reconstruct the whole object complex amplitude in one single snapshot out of focus.The DH capability of applying refocusing in post-processing allows us to track samples in time-lapse experiments.Moreover, the focus can be accurately followed across all the frames of the sequence.We use a DH setup equipped by a 20× microscope objective to follow the time evolution of NIH cells phase-contrast maps when exposed to the above mentioned two concentrations of nGO3.We estimate time changes in biovolume, dry mass and surface area accordingly.Besides, we choose to investigate by FP cells exposed to nGO3 for a deeper study on their interactions on a wider FoV, since in the work reported in [31] the in-vitro cell response resulted spatially inhomogeneous, differently from cell response observed with nGO1 or nGO2.DH observations only allow us to infer information on small groups of cells that can be more or less characteristic of the ensemble behaviour of a larger population inside a Petri dish.In order to add ensemble information, the same population is observed under a custom-built FP microscope.FP usually adopts low magnification, low-NA microscope objectives to yield wide FoV phase-contrast microscopy [40].Lateral resolution is improved beyond the limits of the employed optical system by exploiting a synthetic aperture principle.Samples are probed from multiple angles to collect a set of bright field and dark field images, each one carrying a specific spectral content of the sample [40].The high-resolution phase contrast is estimated after stitching all the low-resolution spectra in the Fourier domain and carrying out an iterative phase retrieval process.Thus, we use FP to inspect the ensemble distribution of the cells over a wide FoV, while attaining single cell level information, and we study the relations between the fate of single cells, the external nGO3 agents, and the whole cells network.It is worth noticing that low magnification optical systems, e.g.bench-top phase-contrast microscopes or even holographic apparatus mounting 4x microscope objectives, could not be used for studying these processes.Indeed, along with wide FoV, the optical system should be able to provide enough lateral resolution to detect small nGO3 spots.Hence, FP is the best candidate to meet these requirements.
The proposed analysis could be extended to monitor different cells, such as tumor and neuronal cells, by DH time-lapse to retrieve quantitative information about cells while uptaking nGO functionalized with drugs.This would be a significant step forward in the new paradigm of personalized medicine, where the drugs are tailored specifically and the fluorescent probes are avoided due to the potential detrimental effects on the cells.In this scenario, we show how the FP method can be complementary to DH for investigating the cell behaviour and its fate in presence of nGO.

Sample preparation 2.1.1. Cell culture
Murine embryonic fibroblasts NIH-3T3 were chosen to monitor by DH time-lapse and FP the effects of nGO in vitro.NIH-3T3 were grown in DMEM supplemented with 10% FBS (both from Life Technologies, Carlsbad, CA), 2 mM L-glutamine (Sigma, St. Louis, MO), 100 U ml −1 penicillin, and 100 µg ml −1 streptomycin.Then, the fibroblasts were seeded at a cell density of 5 × 104 cells ml −1 in a 35 mm Petri dish and incubated at 37 • C and in a humidified 5% CO2 atmosphere in an incubator (Esco).To ensure homogeneous dispersion the nGO3 samples were subjected to sonication with tip for 20 min, and for the double sonication, the mix of nGO3 and DMEM was sonicated in an ultrasonication bath before adding to the cells.In order to investigate the nGO3 cytotoxicity, two concentrations were added to the complete DMEM, 20 µg ml −1 and 100 µg ml −1 .Observations started 24 h after the exposure to the nGO3.Separate Petri dishes were prepared as described for the DH and FP experiments.

nGO preparation/characterization
In this work, attention was focused on the same graphitic material used by Mugnano et al in [31].In particular, EDX, Bright field TEM and DLS analysis were repeated to assess the actual condition of nGO3, in terms of oxidation degree, morphology and size, having been prepared and used for previous works [27][28][29].EDX was performed to quantify the oxidation degree of nGO3, through an evaluation of carbon/oxygen content ratio.EDX analysis was performed using a FEI Quanta 200 FEG SEM equipped with an Oxford Inca Energy System 250 and an Inca-X-act LN2-free analytical silicon drift detector.The sample was obtained by casting the nGO3 dispersion directly on the SEM analysis support.TEM analysis was performed on a FEI Tecnai G12 Spirit Twin (LaB6 source) at 120 kV acceleration voltage.TEM images were collected on a FEI Eagle 4 k CCD camera.Water dispersed nGO3 was collected by immersing TEM copper grids in the aqueous dispersions and dried at room conditions.DLS measurements were carried out using a Zetasizer Nano ZS (Malvern Instruments) at 25 • C and at scattering angle of 173 • to evaluate dHD and (PDI).pH of water dispersions was set at 6.5-7.0 for all the tests.Details about the characterization of the nGO3 can be found in [31].

Digital holographic approach
A custom-built off-axis DH microscopic system has been arranged to analyze NIH 3T3 cells behaviour in terms of morphology and volume changes in presence of different concentrations of nGO3.The details of the DH setup are reported in the supporting material (figure S1(A)).
After time-lapse measurements, the recorded holograms are processed to recover the complex object field (C o ) in the best-focus plane according to the scalar diffraction theory [34]: where ASM is the angular spectrum method [44] to back-propagate the wavefield of the object along the z direction, O F is the demodulated complex image after the selection of the first order of diffraction in the Fourier domain [34,44]; px is the pixel size, λ is the wavelength and d is the best-focus distance.The same process is repeated for the complex reference, as C R (i.e.acquisition without object).The amplitude (A) and the phase (φ) maps are used for the analysis of nGO3 and cells, estimated as follows: where abs refers to the absolute value function and angle calculates the phase from a complex value (arctan function).The ratio between the complex object and the reference object in equation ( 2) has the effect of compensating any wavefront aberration caused by the optics in the system setup [45,46].

Single cells analysis
After obtaining the amplitude and the phase maps of the NIH cells and nGO3 particles in the time-lapse experiments, the amplitude images are used to find the centroids of the nGO3 particles in the cells buffer.A basic threshold-segmentation algorithm is applied to detect the supports of the nGO3 spots that show lower amplitude values than the background.From the nGO3 masks, the centroids of their positions can be calculated.The phase maps are firstly unwrapped (i.e. the phase values are brought back from the [−π, π] range to restore the true phase dynamics of the samples [45]), and used to segment the NIH cells using a watershed-based segmentation algorithm [47].The cells masks help extracting cells morphological information, such as area, major and minor axis lengths, eccentricity and to correctly measure quantitative biophysical quantities such as cells biovolume and dry mass.These will be used to characterize the NIH cells behaviour after adding nGO3 particles.The cells volume is measured as: where Area = N•px•py, with N the number of pixels in the cell support, and px and py are the pixel sizes in the image plane (px = py = 0.1233); CT is the cell thickness in the considered area, which can be estimated considering the phase values of the cells previously retrieved: approximating n c = 1.375 as the cellular refractive index [48] and n m = 1.337 as the refractive index of the medium [49].It is worth pointing out that cells refractive index is considered here spatially uniform within the cell as in [50] since in this analysis we are mainly interested in the changes in time of the biovolume rather than the spatial distribution of the refractive index.Accordingly, we will hereafter neglect the spatial variations of the refractive index.The dry mass is estimated as follows: where α is the refractive increment related to the protein concentration of the cells, here constant (α = 0.2 ml g −1 ) [51].
The area, the biovolume and the dry mass values are calculated for each segmented cell in each frame of the time-lapse recorded video.In each frame, the measured values are averaged over the number of cells present within the FoV to study the trends of the biophysical quantities during the time-lapse experiments, as a function of time.

Fourier ptychographic approach
A custom FP microscope has been built up to retrieve high resolution images over a wide FoV, detailed in the supporting material (figure S1(B)).
Since FP measures only intensity images, to access amplitude and phase values phase retrieval algorithms are needed [52].The phase retrieval iterative process foresees to pass from the spatial domain to the frequency domain and vice versa according to a Gerchberg-Saxton scheme, until converging to a steady optimal value of the complex amplitude of the sample, under certain constraints and hypotheses [40,53].The complex object amplitude (C O,Q ) after Q iterations can be written as follows: where ξ is the functional related to the FP reconstruction algorithm, C O the high-resolution initial guess, and ν the illumination vector [54,55].The FP amplitude and the phase images are obtainable as:

Results and discussions
The EDX analysis allowed us to evaluate the C/O atomic ratio of aged nGO3.It was found to be 2.7 ± 0.6, which is slightly higher than its original value of 2.0 ± 0.2 [27].According to [29,56], GO may undergo spontaneous reduction over time at room temperature, losing a certain number of oxygen-containing groups on the surface of each sheet, giving rise to a higher C/O ratio value.By means of TEM and DLS analyses the shape and size of aged nGO3 particles were evaluated.These characterizations confirmed that the nGO3 morphology was left unchanged, as well as the hydrodynamic diameter, which is around 200 nm, with a PDI equal to 0.35.

DH results and interpretation
Before starting the experiments introducing nGO3 in the cellular environment, we performed a 4 h time-lapse control experiment to check the conditions set of our measurements outside the incubator, which verified the capability of the measurement system to keep the cells alive during the time observation window.Then, we carried out subsequent time-lapse DH experiments for investigating NIH 3T3 cells behaviour and morphology in the presence of nGO3.We recorded first a 3 h time-lapse, with 30 s time step, to study the coexistence of NIH cells and nGO3 at 20 µg ml −1 .Afterwards DH acquisition, we reconstructed the holograms sequence and recovered the corresponding amplitude and phase maps.Successively we performed the same experiments by using the nGO3 at 100 µg ml −1 .The video frames were recorded at 2 min intervals, and we stopped the time-lapse after 4.5 h, after observing cell death.To compare both time-lapse experiments, we set the same experimental conditions, subsampling the first time-lapse with 2 min time step, instead of 30 s time step of the second time-lapse.In the supporting material, some frames of the time-lapse of amplitude maps are shown to report the presence of nGO3 black spots and clusters in the FoV chosen for the time-lapse DH analysis.The segmented nGO3 is reported superposed to the amplitude maps in figures S4 and S5.Details on the nGO3 segmentation process are reported in the supporting material.Figures 1(A)-(C)) show phase maps extracted from the time-lapse sequence at time t = 0 min, t = 88 min and t = 179 min, where it is possible to observe cellular resistance to nGO3.Indeed, the red spots indicate the centroids of nGO3 particles in the medium buffer.However, the phase contrast values and the analysis of the time evolution in the FoV under test indicate that this dose is not lethal for the cells in these experimental conditions.On the contrary, figures 2(A)-(C)) show that cells behaviour has been affected by the presence of nGO3 that promoted cell death.The frames shown in figures 2(A)-(C)) at t = 0 min, t = 88 min and t = 179 min, show clearly the process leading to cell death for all the cells within the FoV, meaning that the 100 µg ml −1 nGO3 concentration inside the medium buffer provokes unhealthy conditions for the NIH cells.In figure S5 of the supporting material, the nGO3 black spots of some frames of the 100 µg ml −1 nGO3 time-lapse experiment are clearly visible from the amplitude maps.These first evaluations find confirmation in the quantitative analysis reported in the following.

Quantitative cells analysis from DH phase images
DH images allowed single-cell quantitative analysis.To analyse NIH fibroblast cells behaviour in presence of nGO3 particles, we evaluated the cellular surface area, the biovolume and the dry mass during the time-lapse experiments.The values of the previous parameters have been estimated after cells segmentation and using equations (3)- (5).Each frame of the time-lapse experiments shows a certain number of cells, and we averaged the parameters frame by frame to assess the average trend within the FoV as a function of time.In figure 1(D), the surface area plot shows that cellular areas tend to increase (with a slope of 1.3) during the time-lapse experiment, confirming the NIH cells coexistence with nGO3 particles at 20 µg ml −1 .The plot of biovolume over time has an oscillating trend until assuming values typical of living cells, providing evident signs of cellular activity, figure 1(E).A polynomial of degree 5 fits well the biovolume distribution, which raises linearly at the end of the experiment.The dry mass plot as a time function has the same slope as the biovolume plot, figure 1(F).Therefore, 20 µg ml −1 of nGO3 do not significantly perturb the cells vitality.Correspondingly, the cells exposed to nGO3 at concentration of µg ml −1 showed no significant internalization of the nGO3 clusters.
In figures 2(A)-(C), the first, the middle and the last phase maps of time-lapse experiment with 100 µg ml −1 nGO3 have been reported that shows the difference in the cell behaviour with respect to figures 1(A)-(C) (20 µg ml −1 nGO3).In figure 2, NIH cells die over time because of nGO3.Results of the DH analysis allowed to quantitatively assess this behaviour.
Figure 2(D) shows at each time point the area averaged over all the cells present in the FoV to estimate the overall trend rather than the fate of a single cell.During the time lapse, cells may enter and/or exit the observed FoV, so that the overall segmented area can increase or decrease accordingly.In this case, we have an increase of the number of cells, and thus the number of active pixels in the image.Nevertheless, during the time-lapse the biovolume decreases, and this would lower the number of active pixels accordingly.The two opposite effects balance each other and the average area in the FoV is barely constant.Moreover, cells surface area tends to slightly decrease during the experiment (with a linear slope of −0.12).Hence, the area measure, averaged over the total number of the cells present in the FoV, shows that this parameter alone cannot help inferring the population fate, while biovolume and dry mass can.Similarly, figure S3) shows the quite constant trend of the area measured over one single cell during the entire time-lapse experiment.The biovolume, figure 2(E), follows a sigmoidal function, which has been overplotted to better show how NIH cells biovolume behaves during the time-lapse measurement with 100 µg ml −1 of nGO3.The decrease of the biovolume in time well describes in quantitative mode the process of cell death.In particular, the sigmoidal function has been estimated as follows: where BV s is the sigmoidal biovolume, BV 0 is the mean value of the biovolumes of the first minutes of the experiment, BV 0 is the mean value of the biovolumes of the last minutes of the experiment (the mean value has been chosen to take into account the variability of the biovolumes due to different adhesion stage for each cell), k is the slope that depends on the constant time and controls the width and the transition area, and t m is the mean time point of the sigmoidal function, i.e. the slope time.The used constant values to determine the sigmoidal function are reported in table 1.The cells clearly try to adapt and survive.In fact, cellular areas and biovolumes change to regulate their physiological mechanisms until the decay time point, where NIH cells start suffering and die at the end of the experiment.Since internalization of the clusters was not observed, cell death could be ascribed in this case to oxidative stress as reported in [32].The dry mass  2(C)).Successively, the cell membrane rupture occurred, and the intracellular fluid flew out the cell body, with a consequent biovolume decrease (see figure 2(E)).The volume decrease of the cells after membrane rupture is a typical sign of a cell death process.Observing figures 1(E) and (F)) and figures 2(E) and (F), we noted two growth peaks of biovolume and dry-mass in different time instants, which correspond to two different cellular events.In figure 3, we zoomed in and analyze the considered time windows.Figure 3(A) refers to a cellular division event, which occurs in the numbered time instants (circled in the zoom).BV and DM describe well in quantitative way the cell division process observable in the sequence of phase maps on the right-side of figure 3(A), which influences the mean value of the parameters in each frame.In figure 3(B), after about 120 min from nGO3 exposure, the cell biovolume initially oscillates according to the regulatory mechanisms that compensate the physiological volume variations to keep an appropriate balance of ions across their cell membrane.The volume data exhibit the upward slope corresponding to the swelling just before the membrane rupture, which is shown in the sequence of phase maps on the right-side of figure 3(B).The intracellular liquid flowed out of the cell; the volume decreased rapidly of about 36% following the trend in figure 2(E).The morphological changes of the cells during nGO3 exposure were evaluated by using another cell parameter, namely the dry mass.The same slope can be observed for dry mass measurements that quantify the total amount of cell mass without considering its water content, and shows a reduction of about 32%, see figure 2(F).
For both time-lapse experiments we found out the coefficients of determination (R-square) of the model data used to fit the biovolume and dry mass distributions to assess the reliability of the observed trends.The biovolume and dry mass in 20 µg ml −1 nGO3 have a R-square value of 0.76, while in 100 µg ml −1 nGO3 of 0.79, which confirm the validity of the fitting models for future prediction outcomes about cells cytotoxicity.Moreover, to highlight the different cells fate in presence of 20 µg ml −1 and 100 µg ml −1 of nGO3, we reported in figure 4 the biovolume and the dry mass trends after 120 min of experiment up to the end of the observation.We linearly fitted the distributions and estimated the corresponding slopes.In the case of 20 µg ml −1 nGO3, the biovolume and the dry mass tend to increase with a slope of 6.08 and 1.155 respectively, which is a sign of cell vitality and activity.On the contrary, in the case of 100 µg ml −1 nGO3 the slopes are −0.214 and −0.0406 respectively, meaning a reduction of cells biovolume and dry mass.The last results enforce the observation of coexistence between NIH cells and 20 µg ml −1 nGO3 and of the lethal effect for 100 µg ml −1 nGO3.In the latter case, the rapid increase of biovolume before the event of cell death suggests a necrosis death event for the cells in the observed FoV.

FP results and discussion
The holographic system does not allow one to gather a comprehensive overview of the health status of the specimen of NIH cells in the Petri dish, since the FoV is not wide enough to cover a large portion of the specimen.In other words, observing only a small FoV over time could be misleading, since the specific conditions expected to determine the cells' fate are in general a function of their distribution in layers or aggregates.To have a broader view of the cellular reaction to nGO3, we acquired the same sample with a FP microscope to obtain high resolution phase contrast images over a 3.3 mm 2 FoV.We considered for this analysis the worst case of the previous DH assay, i.e. 100 µg ml −1 nGO3 concentration.In figure 5, different FP phase contrast images are reported, corresponding to different areas of the Petri dish, to analyse a larger heterogeneity of cells' local density and the corresponding cells' behaviours that can be inferred by such ensemble view.In figure 5(A) we reported a FP phase map where most of cells are alive despite the high nGO3 dose.A possible explanation for this behaviour is the presence of large cellular clusters and dense cellular networks that allow cells enforcing each other and counteracting the nGO3 oxidative stress.This explanation, although interesting and reasonable, should be accurately verified by further future studies.In the inset, we  highlighted the NIH isolated cells (indicated by white arrows) that are probably dead and detached from the dish surface, surrounded by nGO3 particles (yellow spots in figure 5, full FoV image).We observed that isolated NIH cells tend to die, while cells in high density clusters/network are more resilient and able to survive.nGO3 can contaminate and alter the nutrients in the cell medium inducing cells death indirectly.
Moreover, nGO3 may block the delivery of nutrients to cells by adhering to cell membranes and it may also interact with cell membrane surface receptors and can block the transport of various substances into the cell, thus cell death can be also ascribed to cell starvation because of the uptake of nutrients by carbonaceous NPs standing outside the cell membrane [57][58][59].Thus, one possible explanation for the observed trends is that nGO3 clusters interact with the medium buffer thus reducing the amount of cell nutrients available to the cells.Therefore, only the cells interlaced inside dense networks can survive thanks to the collaborative effort of inter-cell transport of material The work in [60] would support this assumption, since it was observed that high cell density monolayers are consistently more resistant to the cytotoxic effects of NPs, compared to sparse monolayers for different cell types.However, modelling and explaining the observed trends would require an extensive specific experimental campaign on larger and more heterogeneous populations and cases, which is far from the scope of this work.Future investigations from our group will be devoted to study comprehensively the link between cells networking and trafficking and their resilience to external stress agents and cues.
The map in figure 5(B) shows several NIH cells arranged in networks that are still alive and stuck to the dish.Again, a number of NIH isolated cells tend to die and take a spherical shape (see white arrows).It is worth pointing out that differences in cells fate as a function of the local cell density are noticeable despite the nGO3 centroids appear uniformly distributed in the FoV.This information on the nGO3 uniformity is an advantageous consequence of accessing wide FoVs and at the same time the high lateral resolution required for segmenting the nGO3 spots.In FP images, thanks to the wide FoV it is possible to observe the quasi-uniformity of the nGO3 spots, thus we exclude any possible effect of locally higher density of nGO3 that could have guided the cells' fate in non-uniform ways inside the large Petri dish.
Here, reactive oxygen species formed on the graphene platelets could have initiated radical reactions, eventually resulting in death for the isolated cells as demonstrated by TCL analysis in the results section of [31,56,61].Figure 5(C) shows a FoV where cells formed lower density clusters and we observed a larger number of isolated cells.The zoomed-in inset shows cells for which the membrane lysis occurred as a consequence of cell death and the cytoplasm was released (indicated by white arrows).This is transparent when probed by visible light but show sharp phase contrast due to the optical phase delay difference with respect to the buffer.A last example showing that NIH cells can survive under a 100 µg ml −1 nGO3 dose if they are interlaced in clusters is reported in figure 5(D).This area corresponds to the central part of the Petri dish, where the cells grew and duplicated more before inserting the nGO3.For this reason, we were able to find denser networks.As shown in the inset zooms, in these cases cells mostly undergo a positive fate despite nGO3 clusters are distributed all around.As mentioned above, along with the possibility of accessing a wide FoV, the FP approach provides QPI information.For each segmented cell, we estimated the mean value of the unwrapped phase values, and we averaged them on the number of cells present in each 500 × 500 patch.Hence, for each patch we obtained a sole average phase value, i.e. the average of the cells mean phase (A-CMP).In this way, we formed a heatmap of the A-CMP values, in order to analyse the global content of the FP image at a fixed scale (500 × 500 in this example).In figure 6(A), we superimposed the A-CMP heatmap on the FoV in figure 5(C), where NIH cells appear to be mostly subject to cytoplasm extrusion.The red spots are the centroids of cells, whose segmentation borders are in light blue.The black spots represent the nGO3 particles, segmented from the amplitude maps of the FP images.Yellow colour of the A-CPM heatmap patches indicates high phase values, which can refer to a higher number of cells present in the patch or to a particular cellular state (duplication or death).Blue colour means absence of cells in the patch.The same approach has been applied to generate a heatmap of the standard deviation of the background phase (STD-BP) values patch by patch.The standard deviation, which measures the dispersion of the phase values in the patch, is here used to measure the presence of material extruded from the cells into the background medium.Yellow colour of the STS-BP heatmap patches means higher dispersion of cellular extrusion material, while the blue colour indicates lower dispersion.A further QPI-FP result is reported in figure 7, where we evaluated the area, the biovolume and the dry mass of NIH cells of the FoV in figure 5(C).For each segmented cell, we estimated the mean value of the area, biovolume and dry mass, and we averaged them on the number of cells present in each 500 × 500 patch.We calculated histograms of the previous estimated values, for all patches of the full FoV.The QPI-FP parameters are compared to the QPI-DH parameters in figures 2(D) and (F), which has the same nGO3 concentration (100 µg ml −1 ).In figure 7(A), the area values are mostly ranged in [106, 400] µm 2 (excluding 0 values), as the QPI-DH area values shown in figure 2(D).In figure 7(B), the biovolume values of most patches are in a range of [187, 1000] µm 3 (excluding 0 values), i.e. in good agreement with the QPI-DH biovolume values measured in figure 2(E).In figure 7(C), the dry mass values mostly range in [36,150] pg (excluding 0 values), in good agreement with the QPI-DH dry mass values reported in figure 2(F).Noteworthy, histograms are meaningful in the case of FP since it allows accessing a wider FoV, unlike the case of DH images where the number of cells is limited by the FoV.The substantial agreement with the two methods is a further confirm of the validity of the multi-sensor approach for characterizing in quantitative mode the cells fate and coexistence with external agents.In this sense, the use of FP-QPI is very promising since it allows a multi-scale investigation of NIH cells over a wide FoV and a resolution large enough to detect the nGO3 tiny spots.Results presented here also highlight the importance of observing large areas of the wells where cells are seeded.Indeed, their behaviour in the presence of external agents can be highly heterogeneous and small FoV observations could lead to erroneous interpretations.

Conclusions
Assessing cell tolerance to the presence of nGO and understanding the amount of nGO that makes the cell environment toxic or unsustainable in terms of cells growth is a common goal in the field of nanomedicine.Here, we have proposed a multimodal/multiscale quantitative analysis of NIH 3T3 fibroblast cells in presence of nGO3 particles in two different concentrations by means of DH and FP.The DH phase images allowed us performing time-resolved single cell assays with high resolution interferometric microscopy in transmission mode, quantitatively evaluating the biovolume and the dry mass of the NIH 3T3 cells in the presence of 20 µg ml −1 and 100 µg ml −1 nGO3 concentrations, respectively.Results showed that the 20 µg ml −1 concentration of nGO3 does not perturb the cell equilibrium and does not significantly guide the cells fate, while a 100 µg ml −1 concentration of nGO3 can lead to cell death despite nGO3 form large clusters that are not internalized by the cells.However, we observed that this high resolution, small FoV assay alone is not sufficient to offer a clear comprehensive view of the phenomenon.On the other hand, FP images gave us the possibility to analyze the well over a wide FoV with single cell level lateral resolution (0.5 µm resolution over a 3.3 mm 2 FoV for the setup employed in our experiments).The ensemble view allowed us correlating the cells fate to the distribution of nGO3 clusters and, above all, to the cell local density.The FP results suggested that isolated cells tend to die with higher occurrences, while the cells cluster are able to counterbalance the adverse effects of nGO3 even for the lethal dose.A possible explanation, discussed in the previous section, is that non internalized nGO3 clusters interact with the medium buffer and subtract nutrients to the cells, impairing their normal life cycle.Cells in denser aggregates can enforce each other in a collaborative effort in which they exchange biological material.This hypothesis should be verified by further analysis, e.g.adding a parallel imaging channel relying on fluorescence.At this stage, our work provides a quantitative multimodal test that shows how DH can be complementary supported by FP results to gather information at the single cell level (DH) and the ensemble cell network level (FP).Further improvements of our test methodology will involve the increase of time resolution for the DH analysis and to implement time resolved FP assays, e.g. using fast multiplexed illumination schemes.

Figure 1 .
Figure 1.Time-lapse 1. NIH 3T3 cells and 20 µg ml −1 nGO3 observation.(A) Phase map of NIH cells at the beginning of the experiment, the colorbar refers to phase values in a range of [0, 3.5] radians; (B) phase map of NIH cells in the middle of the experiment, the colorbar refers to phase values in a range of [0, 3.5] radians; (C) phase map of NIH cells at the end of the experiment, the colorbar refers to phase values in a range of [0, 3.5] radians; (D) cells surface area plot as time function; (E) cells biovolume plot as time function; (F) cells dry mass plot as time function.The red spots represent the graphene localization in the surrounding environment.

Figure 2 .
Figure 2. Time-lapse 2. NIH 3T3 cells and 100 µg ml −1 nGO3 observation.The time-lapse experiment.(A) Phase map of NIH cells at the beginning of the experiment, the colorbar refers to phase values in a range of [0, 6] radians; (B) phase map of NIH cells in the middle of the experiment, the colorbar refers to phase values in a range of [0, 9] radians; (C) phase map of NIH cells at the end of the experiment, the colorbar refers to phase values in a range of [0, 5] radians; (D) cells surface area plot as time function; (E) cells biovolume plot as time function; (F) cells dry mass plot as time function.The red spots represent the graphene localization in the surrounding environment.

Figure 3 .
Figure 3. Difference among the growth peaks in biovolume and dry mass in the two time-lapse experiments.(A) NIH 3T3 cells and 20 µg ml −1 nGO3: the growth peak zoomed in the middle part shows a cellular division, visible in the sequential phase maps on the right-side (the colorbar refers to phase values in a range of [0, 3.5] radians); (B) NIH 3T3 cells and 100 µg ml −1 nGO3: the growth peak zoomed in the middle part shows a cellular death, visible in the sequential phase maps on the right-side (the colorbar refers to phase values in a range of [0, 9] radians).The circled number is the value of the parameter obtained averaging the values measured for all the cells in the FoV.

Figure 4 .
Figure 4. Time-lapse comparison: Left-side) Biovolume plot for both time-lapse experiments, i.e. 20 µg ml −1 and 100 µg ml −1 nGO3 concentrations, after the first 120 min until the end of experiments.The values have been fitted to evaluate the slopes of biovolume distributions.Right-side) Dry-mass plot for both time-lapse experiments, i.e. 20 µg ml −1 and 100 µg ml −1 nGO3 concentrations, after the first 120 min until the end of experiments.The values have been fitted to evaluate the slopes of dry-mass distributions.

Figure 5 .
Figure 5. FP image analysis.(A) FP phase contrast image of NIH cells and nGO3.The inset zooms the dead NIH isolated cells (indicated by white arrows), and the alive cellular clusters; (B) FP phase contrast image of a less dense distribution of NIH cells and nGO3.The inset zooms the dead NIH isolated cells, while the denser clusters have higher probability to survive; (C) FP phase contrast image of a less dense distribution of NIH cells and nGO3.The inset zooms the death of NIH isolated cells, which break their membranes; (D) FP phase contrast image of a dense distribution of NIH cells and nGO3.The inset zooms the NIH cells in cluster, which are still alive despite the nGO3 in the surrounding environment.The nGO3 is reported with red spots over the entire FP FoV, while it is shown with black spots in the insets.The colorbar refers to phase values in a range of [−4, 4] radians.

Figure 6 .
Figure 6.QPI information from FP phase maps: (A) A-CMP heatmap of NIH cells superimposed on the FoV of figure 5(C); (B) STD-BP heatmap superimposed on the FoV of figure 5(C).The red spots are the centroids of NIH cells, contoured in light blue.The black spots are the nGO3 particles.Each patch is ∼105 µm.Colorbar values are in radians.The scale bar is 500 pixels, i.e. ∼105 µm, for both images.

Figure 7 .
Figure 7. QPI-FP parameters referring to FP FoV of figure 5(C): (A) histogram of area mean values of NIH cells patch by patch; (B) histogram of biovolume mean values of NIH cells patch by patch; (C) histogram of dry mass mean values of NIH cells patch by patch.

Table 1 .
Constant values used in the sigmoidal function for fitting both biovolume and dry-mass distributions.In particular, k = 0.5 is a sigmoidal coefficient that allows to control the width of the transition area to better fit the values estimated.The value tm is the central time of the sigmoidal slope (114 min).The initial values of both biovolume and dry-mass have been estimated considering the mean value of the initial time point.The final values have been estimated considering the mean value of the final time point.The initial and final dry mass values used to estimate the sigmoidal fitting are reported in table 1.Cells exposed to 100 µg ml −1 of nGO3 showed a morphology change from elongated to spherical shapes.Such morphological variation is typically observed before cell death.Moreover, evident features of low vitality appeared with fibroblasts exhibiting cytoplasmic shrinkage and/or blebbing.As further evidence, detaching from each other or floating in the medium was observed (see figure