Nominally identical microplastic models differ greatly in their particle-cell interactions

Due to the abundance of microplastics in the environment, research about its possible adverse effects is increasing exponentially. Most studies investigating the effect of microplastics on cells still rely on commercially available polystyrene microspheres. However, the choice of these model microplastic particles can affect the outcome of the studies, as even nominally identical model microplastics may interact differently with cells due to different surface properties such as the surface charge. Here, we show that nominally identical polystyrene microspheres from eight different manufacturers significantly differ in their ζ-potential, which is the electrical potential of a particle in a medium at its slipping plane. The ζ-potential of the polystyrene particles is additionally altered after environmental exposure. We developed a microfluidic microscopy platform to demonstrate that the ζ-potential determines particle-cell adhesion strength. Furthermore, we find that due to this effect, the ζ-potential also strongly determines the internalization of the microplastic particles into cells. Therefore, the ζ-potential can act as a proxy of microplastic-cell interactions and may govern adverse effects reported in various organisms exposed to microplastics.


Supplementary Notes
Supplementary Note 1: Particle SEM micrographs The size and shape of the microplastic particles were investigated by using scanning electron microscopy (SEM).Most microplastic particle types deviated by less than 3% from the nominal diameter specified by the manufacturers.However, TJ, and PX deviated by 12% and 4% (Table 1).Furthermore, we found that the surface morphologies of the pristine microplastic particles show clear differences between the particle types (Figure 1, Supplementary Table 1).ST and TJ particles had the most irregular surfaces compared to all other particles.We observed indentations on the surfaces of the TJ particles and elevations on the PX and TJ particles.The other particles displayed in Figure 1 (MM-C, MM, MG, KI and TS) were highly spherical with smooth surfaces, although MG seemed to be covered by a net-like structure.The environmentally exposed particles (MM-SW2 and MM-SW4, MM-FW2, and MM-FW4) were coated heterogeneously by biomolecules and organic debris (eco-corona) as shown previously 1 and further analyzed by STXM and XPS measurements (Supplementary Figure 4, Supplementary Tables 2 and 3).The average diameter of MM-SW2, MM-SW4, and MM-FW4 increased slightly by a few nanometers, indicating a thin layer of biomolecules 1 .Furthermore, the standard deviation of the mean diameter of all environmentally exposed microplastic particles substantially increased, indicating that they became less monodisperse.Accordingly, also their eccentricity and surface roughness increased.

Supplementary Note 2: Validation of the flow profile
To validate that the derived velocity profile  x,par (, ) given by equation (15) matches the flow field, which is established in the channels, we measured the velocity profile inside the channel with a high-speed camera and particle tracking.A small number of carboxylated polystyrene MPs with a diameter of 1 μm (micromer, micromod Partikeltechnologie GmbH, Rostock, Germany) was dispersed in MilliQ water at a concentration of approximately 2 × 10 5 µL -1 .The solution was pumped through the channel at flow rates between 0.2 µL s -1 and 2 µL s -1 , which was controlled by setting the motor velocity according to equation (8).The resulting flow field  x,par (, ) (equation ( 15)) was measured on an inverted, motorized microscope (Nikon Eclipse Ti, Nikon, Tokyo, Japan) with a 40x water immersion objective with a high numerical aperture to minimize the focal depth (CFI Apo LWD 40× WI λS, Nikon, NA = 1.15).Brightfield images of the flow at various  and  positions were captured with a high-speed camera (IDT Nx4-S2, Integrated Design Tools, Pasadena, California) at frame rates of up to 2000 Hz.
To detect the particles automatically, the images were filtered with a spatial band pass filter in order to reduce noise and to remove any signals on length scales larger than the particles.Secondly, the images were filtered with a moving median filter in time to single out all stationary objects, e.g.sensor dust.These filtered images were then subtracted from the images.The moving particles were then detected with a custom-written peak-finding routine and tracked with a particle tracking algorithm 2 .The code was implemented in MATLAB 2019b.For every position inside the channel, the velocity of typically 50-100 particles was averaged.Slight deviations from the mean value could be attributed to particles located in slightly lower or higher  positions than the main focal plane, which were still sharp enough to be tracked.The measured flow profile  x (, ) was in agreement with the theoretical prediction (Supplementary Figure 11).
As we used different flow rates during the rupture experiments, we also tested whether the velocity inside the channel scales linearly with the flow rate as expected from equations ( 14) and ( 15).The high-speed camera was fast enough to capture particles with velocities of up to 3.5 mm s -1 at a flow rate of 2 µL s -1 with sufficiently short frame time intervals for the particle tracking algorithm.Up to 2 µL s -1 , we found perfectly linear scaling, demonstrating that the pressure even at relatively low flow rates is high enough to properly close the check valves in the tubing system (see Supplementary Figure 12).During the calibration experiments and during the cell experiments, we observed straight particle trajectories far away from the cells, indicating laminar flow in the channels.
Since the measured velocity profile  x (, ), both in shape and magnitude, was in excellent agreement with the theoretical prediction (see Supplementary Figures 11 and 12), we concluded that equation ( 17) can be used to correct for the parabolic height profile.
Supplementary Figure 1: ζ Supplementary Figure 4: STXM analysis of the eco-corona.C1s NEXAFS spectra extracted from Scanning Transmission X-ray Microscopy (STXM) datasets from regions directly at the surface of PS beads that were previously incubated in seawater, freshwater and deionized water respectively.The surface regions, highlighted in red, were selected in the STXM images based on the average optical density (OD) range of 0.1-0.9across the C1s absorption edge, which is equivalent to a cumulative thickness of up to 100 nm, arranged tangentially around the PS beads.These regions are marked in red in the respective STXM images.The equivalent thickness was calculated using the atomic scattering factors 3 , the formula C8H8 and an assumed density of 1.09 g cm -3 of the polystyrene.All 3 spectra were decomposed into a sum of individual gaussian peaks plus the ionization edge modelled as an arctan function.A minimum of 7 analytical peaks was required and used for fitting the respective spectra: 284.0 eV (quinone C=O), 285.0 and 285.4 eV (aromatic C=C), 287.4 eV (aliphatic C-C), 288.2 eV (protein C-O), 288.9 eV (carboxylic C-O), 289.5 (polysaccharide C-O).Peak energies and widths were optimized and fixed at the same values for all 3 spectra, whereas the respective peak areas were fitted (Supplementary Table 2) using the peak fitting algorithm of Athena 4 .For illustrating the peak energies that are subtle only in the spectra of the thin regions of the eco-corona around the PS bead, 3 reference materials are presented as well, for protein (albumin), acidic polysaccharides (alginate) and neutral polysaccharides (xanthan) 5 .The dashed grey lines indicate the respective peak energies of the gaussians used for fitting.The data lines of the spectra were offset for clarity.

Supplementary Figure 5: Binding kinetics of microplastic particles to coverslips.
Binding kinetics to coverslips significantly differed between samples (kon: Kruskal-Wallis test, two-sided P = 8.69 × 10 -16 ; koff: Kruskal-Wallis test, two-sided P = 5.77 × 10 -10 ; Irreversible binding: Kruskal-Wallis test, two-sided P = 2.83 × 10 -6 ).We measured rates of a similar magnitude as for the binding of microplastic particles to cells.This means that particles which strongly bound to cells also bound strongly to coverslips.However, particle-coverslip adhesion was in general slightly weaker than particle-cell adhesion (compare Figure 2).This was reflected by a generally lower kon (a) to coverslips, a higher koff (b) from coverslips, and a generally lower fraction of irreversible binding events (c).Like for the binding to cells, the binding kinetics were strongly correlated to the particles' ζ-potential.With increasing negative ζ-potential kon increased (Pearson's R = 0.9, two-sided P = 9.4 × 10 -6 ), koff decreased (Pearson's R = -0.8,two-sided P = 0.003), and the fraction of irreversible binding events increased (Pearson's R = 0.8, two-sided P = 0.003).In all panels, error bars represent standard error of mean of n = 9 measurements (for each measurement, on average 550 particles were analyzed).Source data are provided as a Source Data file.
Supplementary Figure 8: Internalization and maturation of all particle types.All particle types underwent a similar form of internalization and maturation: After the particles interacted with the cells, there was a substantial increase in the LifeAct signal at the site of the particles, indicating that filamentous actin was polymerized.This shows that particles were internalized by an actin-dependent pathway, such as phagocytosis or macropinocytosis.After the LifeAct signal around the particle decayed, indicating depolymerization of the actin filaments, the LysoTracker signal gradually increased over time.This shows that internalized particles interacted with lysosomes and became acidic, undergoing a maturation process that is typical for phagosomes or macropinosomes.Source data are provided as a Source Data file.velocity field given by equation ( 15) was fitted to the measured data with the flow rate in the center as the only free parameter.The measured velocity field matched the velocity field  (x,par) calculated from the motor velocity and the channel geometry within the margin of error.The measured flow field inside the channel matched the flow field predicted from equations (8), (14), and (15) within the margin of error, which was calculated by gaussian error propagation.Notably, this was also the case near the side walls of the channel, where the channel was approximately 15 µm thinner.Source data are provided as a Source Data file.Statistical testing of the relative attachment between particles and cells as well as between particles and coverslips.Particles of different manufacturers attach significantly different to cells as well as coverslips.The data was tested for a normal distribution using a Shapiro-Wilk test and for a homogeneity of the variances with a Levene test.Differences between particles were identified with a Kruskal-Wallis test together with a Games-Howell post-hoc test.Different letters denote groups of particle types between which significant differences with a two-sided P < 0.05 were detected.For example, kon to cells of PY was statistically different to MM, MG, KI, TS, PX, MM-FW2, MM-FW4, MM-SW2, and MM-SW4, but no statistically significant differences existed between PY and ST and TJ.A detailed summary of all statistics is presented in Supplementary Data 1.

Supplementary Figure 3 :
-potential before and after incubation in cell culture media.The ζ-potential of microplastic particles after an incubation of 2 h in cell culture media ζmedium was strongly correlated to their initial ζ-potential ζinitial (Pearsons's R = 0.8, P = 0.004).Detailed values of the ζ-potential are given in Supplementary Table 1.Source data are provided as a Source Data file.Organisms in environmental media.Various microorganisms were present in the environmental media.(a) and (b) show examples of freshwater organisms in brightfield and fluorescence microscopy (Texas Red channel, exc.542-582 nm, em.604-644 nm; color scales represent fluorescence in arbitrary units).(c) and (d) show examples of saltwater organisms in brightfield and fluorescence microscopy.(e) and (f) show scanning electron micrographs of freshwater organisms.(a) possibly shows a diatom.(b) and (f) possibly show green algae of the genus Lagerheimia.(c), (d), and (e) possibly show examples of cyanobacteria.The autofluorescence of the microorganisms in the Texas Red channel might be caused by chlorophyll.