Characterizing the Transport and Surface Affinity of Extracellular Vesicles Isolated from Yeast and Bacteria in Well-Characterized Porous Media

Extracellular vesicles (EVs) are membrane-bounded, nanosized particles, produced and secreted by all biological cell types. EVs are ubiquitous in the environment, operating in various roles including intercellular communication and plant immune modulation. Despite their ubiquity, the role of EV surface chemistry in determining transport has been minimally investigated. Using the zeta (ζ)-potential as a surrogate for surface charge, this work considers the deposition of EVs from the yeast, Saccharomyces cerevisiae, and two bacterial species, Staphylococcus aureus and Pseudomonas fluorescens, in well-characterized porous medium under various background conditions shown to influence the transport of other environmental colloidal particles: ionic strength and humic acid concentration. The affinity of S. cerevisiae EVs for the porous medium (glass beads) appeared to be sensitive to changes in ionic strength, as predicted by colloid stability (Derjaguin, Landau, Verwey, and Overbeek or DLVO) theory, and humic acid concentration, while P. fluorescens EVs deviated from DLVO predictions, suggesting that mechanisms other than charge stabilization may control the deposition of P. fluorescens. Calculations of attachment efficiency from these deposition studies were used to estimate EV transport using a clean-bed filtration model. Based on these calculations, EVs could be transported through such homogeneous porous media up to 15 m.


INTRODUCTION
−3 As cells are ubiquitous in almost every environmental compartment, so are their EVs.From eliciting plant immune responses 4 to scavenging nutrients for their parent cells, 5,6 EVs can mediate intercellular and interorganismal communication in the environment, properties integral to microbial systems and the ecosystems in which they exist.Yet, despite their importance, EV environmental transport has been only minimally explored and for a very limited series of organisms.Bos et al. tracked the diffusion of Escherichia coli-derived EVs using fluorescence microscopy, demonstrating that these EVs tended to remain close to the parent cells when the parent cells were not under stresses such as extreme environmental conditions. 7The examination of mesenchymal stem cell-derived EV transport across a hydrogel barrier by Lenzini et al. showed that these EVs were released from hydrogels at a higher rate than liposomes of a similar size and lipid content. 8Only a few studies have examined long-range EV transport from a colloidal perspective, especially regarding EV deposition.To address this shortcoming, this work seeks to examine EV transport patterns experimentally and quantifiably for three microbial EVs using methods and theory drawn from traditional colloidal studies.
The recent use of classical colloidal analysis techniques to investigate EV transport has begun to yield insight into the physical−chemical factors contributing to EV dispersion and stability in the environment.For example, reports characterizing the ζ-potential of human choriocarcinoma cell EVs noted that the EV ζ-potential was sensitive to changes in pH and the presence of multivalent ions but largely unaffected by increasing concentrations of monovalent ions. 9Additionally, we previously evaluated the ζ-potential of EVs from three different microbial organisms in varying environmental conditions, demonstrating that in most cases, the ζ-potential of EVs was largely unaffected by changes in ionic strength but was sensitive to pH and humic acid (HA) concentrations. 10It should be noted that these characteristics (e.g., their ζ-potentials being affected by pH or the presence of ions or humic acid in certain concentration ranges) are common for colloids, which supports the treatment of EVs as colloids.Significant differences were also found between the ζ-potentials of the bacterial EVs and those of their respective parent cells.From a study of E. coli EVs, Gnopo et al. determined that acidic pHs and high ionic strength increased EV aggregation. 11While these studies provide an initial glimpse into the conditions that likely impact interactions between EVs and surfaces in the environment, predicting their transport in the environment will be challenging unless the surface properties of EVs are more thoroughly interrogated.
To characterize the transport of colloidal particles, researchers have refined methods to quantify particle attachment to surfaces.−14 The parameter α may take on values between 0 and 1, describing the probability of attachment between two surfaces upon contact and typically interpreted as being a function of particle surface properties and the local environmental conditions.The attachment of particles to surfaces described by α differs from the phenomena of adsorption and partitioning, which assume thermodynamic equilibrium between particle adsorption and desorption.The complexity of environmental and physiological systems currently require the experimental determination of α rather than being calculated from colloid stability theory, such as Derjaguin, Landau, Verwey, and Overbeek (DLVO) theory. 15,162][13][14]29,30 By applying a similar approach to EVs, our goal here is to obtain estimates of a "sphere of influence" that a cell may exert in the distribution of EVs in an environmental setting. For eample, the distance across which EVs from a microbe might reach plant root systems in agriculture or the maximum transport distance that EVs carrying nutrients could be delivered in a groundwater system could be predicted by means of α experiments and modeling.Such estimates however must be treated as relative indicators due to the influence that heterogeneity may have in an actual environmental setting.
Relying on the success of column studies to characterize particle deposition, the goal of this work is to characterize α for EVs under various conditions of solution chemistry, specifically changes in ionic strength and the presence of a naturally occurring organic matter, HA.These environmental factors were specifically chosen to allow this work to build on previously published ζ-potential data for EVs under similar environmental conditions. 10These evaluations allow for a comparison between observations of the ζ-potential and trends in attachment efficiency, α, and consequently inform a deeper understanding of the role electrostatics play in EV deposition.The values of α obtained from these column studies were used to determine the collector efficiency of the EV and a glass bead collector 17,31 in a clean-bed filter model. 32,33Finally, the collector efficiency was used to predict EV transport through a hypothetical column of saturated porous media where the relative influence of HA concentration and ionic strength on EV transport can be compared.Collectively, α values and the associated model will allow for initial predictions for transport distances of EVs in the environment and highlight conditions that might facilitate or prevent further transport.

Organisms, Growth Conditions, and Reagents.
The three model organisms used in this study were grown in liquid cultures to a stationary phase with shaking, as previously described in Rogers et al. 10 4,34 To remove cells, cultures were centrifuged (Eppendorf 5804R, Rotor: FA-45-6-30; 10,000g, 30 min, 4 °C).Supernatants were then filtered with a 0.45 μm poly(ether sulfone) (PES) filter to ensure that all cells were removed.EVs were then concentrated using tangential flow filtration with a 100 kDa cutoff filter (Pall Corporation, Port Washington, NY) and then filtered again using a 0.45 μm PES syringe filter (Pall Corporation, Port Washington, NY, P: 60206).Finally, EVs were pelleted via centrifugation (Beckman Optima TLX ultracentrifuge, Rotor: TLA-100.3;90,935g, 1 hr, 4 °C) and resuspended in 50 mM PBS.EV samples were stored at 4 °C and analyzed shortly thereafter.

EV Quantification by the Bradford Assay and Sample Preparation.
Ten microliters of each of the final EV stocks was added to 300 μL of Bradford reagent (VWR, Radnor, Pennsylvania) in duplicate, and the absorbance was read at 595 nm.To determine the protein concentrations of the EV stocks, these readings were compared with the standard linear curve generated using a series of bovine serum albumin concentrations.To prepare samples for particle imaging and analysis, EV stocks were diluted with deionized water to a protein concentration of 2 μg/mL and supplemented with the appropriate amount of 1× PBS and/or humic acid (HA) solution to adjust to the desired levels.A stock of 50 mg/L Pahokee Peat standard humic acid (International Humic Substance Society [IHSS], cat.#1S103H) was used as the HA solution, which is a standard agricultural soil humic acid, following a similar protocol to Rogers et al. 10 The initial pH value was between 6 and 8, depending on initial solution Environmental Science & Technology composition and was adjusted to pH 7.0 ± 0.2 using 0.1 M HCl or NaOH with mixing.This addition of acid or base was determined to be less than 10% of the total ionic strength and follows a similar protocol to Rogers et al. 10 2.4.Size and ζ-Potential Evaluation.Size determinations were performed using a Malvern ZetaSizer ZS (Malvern, U.K.), as in Rogers et al., and a transmission electron microscope (TEM).The light intensity distribution from the dynamic light scattering (DLS) data comes from nine measurements (triplicate measurement of three biological replicates).EVs were prepared for TEM imaging by deposition on formvar/ carbon grids (400 mesh, copper).15 μL of sample was pipetted onto the grid, left for 20 min, and the remaining liquid was gently wicked away.Then, 5 μL of vanadium negative stain (Abcam, ab172780, Cambridge, U.K.) was placed on the grid and left for 10 min, and the remaining stain was wicked off.The grid was rinsed with 15 μL of nanopure water and air-dried for 2 h.TEM imaging was performed using an JEOL TEM 2100 Plus microscope (JEOL; Japan) with a controlled exposure time between 7 and 8 s.Images were then processed for size using the NIH's Image J software.
Electrophoretic mobility measurements (which were then converted to the ζ-potential using the Henry equation 35 ) were conducted also using the same Malvern ZetaSizer as was used for the size measurements.The sample settings for refractive index (1.330) and absorption (0.060) for liposomes were used for ζpotential calculations.The initial pH value was between 6 and 8, depending on the initial solution composition, and was adjusted to pH 7.0 ± 0.2 using 0.1 M HCl or NaOH with mixing.Standard error is reported to compare the mean values of the ζpotential between the different organisms and their conditions.
2.5.Glass Bead Preparation.Spherical glass beads were used as porous medium in this work.To prepare glass beads, methods from Pelley and Tufenkji were followed. 36In this method, 1 L batches of 360 μm glass beads were soaked in 12N HCl for 24 h.After acid washing, the beads were rinsed thoroughly with deionized water until the pH of the decanted liquid from the washing was at least 5.6 to ensure that all acid was removed from system.Beads were transferred to crucibles and baked for an hour at 120 °C to evaporate all water.Beads were then baked at 600 °C for 5 h to remove all organic content from the surface of the packing materials.The cleaned materials were stored in a clean, airtight container until use.

Column Description and Operation.
A schematic of the column apparatus is provided in Figure S1.Briefly, the column testing setup was arranged as follows: 13 g ± 0.1 g of dry glass beads were packed into a glass column (10 mm × 150 mm, Diba Omnifit EZ Chromatography Column, Cole Parmer).Deionized water was pumped into the column from the bottom by means of a syringe pump until the whole column was filled with water.The column effluent was fed into an in-line ultraviolet−visible (UV−vis) spectrophotometer (Evolution 201 UV−visible Spectrophotometer, CAT# 912A0883, Thermo Fischer Scientific), and the spectrophotometer effluent was discarded.Roughly 10 pore volumes of the background electrolyte (i.e., PBS or PBS with HA) was passed through the column to equilibrate the packing material with the electrolyte, following established protocols. 14,37The spectrophotometer was blanked to the background electrolyte.Then, the EV sample was passed directly into the spectrophotometer, bypassing the column by means of T-connectors, to obtain a maximum absorbance at 280 nm for the sample (the initial EV concentration, C 0 ).The tubing was then flushed with more background electrolytes to remove all of the EV sample from the system.The EV sample was then passed directly through the column at a flow rate of 0.8 mL/min, and the absorbance at 280 nm was recorded at approximately every 10 s for at least four bed volumes (i.e., until the volume passed through the packed bed was equal to 4 times the volume of the packed bed).Experiments where EV type, ionic strength, and HA concentration were varied were conducted at a flow rate of 0.8 mL/min, which corresponds to a Darcy velocity of 0.017 cm/s.
A tracer study using potassium nitrate was performed to evaluate the integrity of the column and the residence times of associated connections (Figure S2).Tracer tests verified the residence time of the column and indicated that the system plateaued (i.e., the signal demonstrated that the packed bed could no longer "collect" more particles) after approximately 2.25 bed volumes, indicating the clean-bed volume where the peak C/C 0 values for the model were to be taken.The details of how the tracer study was performed are included in the Supporting Information.

Calculation of Attachment Efficiency (α).
Classic filtration theory was used as the framework for interpreting particle removal by the saturated porous medium in terms of the attachment efficiency.Particle concentration as a function of the length of the porous medium is derived from a mass balance across the medium, with deposition occurring as a first-order decay with respect to the distance 38 where C is the concentration of particles (EVs), z is the distance through the porous medium, and λ is the first-order deposition rate constant or filtration coefficient.This equation can be integrated by using a relationship for λ, which is expressed as a function of the single collector efficiency, η 0 , the attachment efficiency, the diameter of a single collector (e.g., a single glass bead), and the porosity of the porous medium. 23Substituting for the filtration coefficient and rearranging the attachment efficiency can be expressed as a function of the experimentally observed concentrations of particles in the porous medium effluent, C and influent C 0 i k j j j j j y { z z z z z i k j j j j j y where d c is the collector (glass bead) diameter, ϵ is the porosity, and L is the length of the bed.η 0 can be calculated from the various parameters for the system based on the equations provided in Tufenkji and Elimelech; 17 tabulated parameters used in these calculations are listed in Table S1.The value of α was determined using eq 2, with the value of C/C 0 taken from the steady-state phase of the column data.Experiments for α were conducted in duplicate, and standard error is reported to compare between different experiments.

Model Validation.
To predict the transport distance of EVs, eq 2 was rearranged to solve for the column length i k j j j j j y { z z z z z i k j j j j j y Maintaining α as a constant, eq 3 was solved for a variety of lengths L, across which an EV may be transported.Eq 3 was solved analytically in MATLAB, R2019a 64-bit academic use license, on a 2018 MacBook Pro laptop.The MATLAB code and Environmental Science & Technology a full table of collector efficiency parameter values can be found in the Supporting Information.
To validate the clean-bed filtration model used as the basis for calculating attachment efficiency for these vesicles, additional column experiments were performed at different flow rates, specifically at 0.2, 0.4, and 0.8 mL/min.These comparisons were only performed for S. cerevisiae EVs in 1 mM PBS and without HA, which correspond to the same parameters inputted into the model.
2.9.Statistical Analysis.Statistical analysis was conducted for both the ζ-potential and C/C 0 results as a function of media conditions and the organism from which the EVs were

Environmental Science & Technology
produced.To test for significance, one-way analysis of variance (ANOVA) and subsequent pairwise testing with Tukey's honestly significant difference (HSD) test was performed in MATLAB for each organism's EVs and for both changes in ionic strength and HA concentration separately.The same statistical test was performed for C/C 0 results for different Darcy velocities, using one-way ANOVA and Tukey's HSD tests.The tabulated results of the ANOVA and Tukey's HSD tests are included in the Supporting Information.

RESULTS AND DISCUSSION
3.1.Characterization of EVs.EVs were collected from the pelleted fraction of culture supernatants from S. cerevisiae, S. aureus, and P. fluorescens.The size of the EVs was determined by DLS and TEM to both validate the presence of EVs and to identify any difference in size that may influence transport (Table 1).The light intensity distribution from the DLS data is presented in Figure S5.TEM images confirmed the presence of EVs (Figure S6), showing similar features to previous reports. 39,40The ranges of ionic strength and humic acid (HA) were chosen to reflect conditions commonly found in the environment and to compare with previous results. 10s shown in Table 1, measurements using TEM of all three types of EVs revealed a lower measured particle diameter compared to the measurements with DLS.This is consistent with previous research, in particular with TEM tending to underestimate EV size while DLS tending to overestimate EV size. 41In addition to these size measurements, the TEM images show that there is a coating on the S. aureus EVs (Figure S6c).It is likely that this coating is a concentrated extracellular material from the tangential filtration step of the EV preparation methods.While this material may associate with the EVs upon dehydration, this material may be contributing to the size distribution of size measurements within the samples.

Breakthrough Curves and α Determination.
To determine the transport capabilities of EVs, column analyses were performed for all three EV types using various ionic strength and HA conditions (Figure 1).The curves shown here are typically referred to as "breakthrough curves", where the analyte (EVs, in this case) becomes detectible in the effluent.Here, EVs are first detected around 1 bed volume (the point of "breakthrough") and are detected until an equilibrium between deposited particles onto the packed media and the suspended particles is reached.Experimentally, this equilibrium is observed when the normalized concentration reaches a plateau with respect to time.While not all breakthrough curves indicate this equilibrium (i.e., there is no plateau), the equilibrium point is determined by a tracer experiment (Figure S2), and for this study, the equilibrium point was reached at 2.25 bed volumes.The normalized concentrations for all breakthrough curves at this plateau point are shown in Tables 2 and 3.
According to ANOVA, only the C/C 0 values for S. cerevisiae EVs were affected by changes in ionic strength.Tukey's HSD test showed that the plateau for S. cerevisiae EVs in 1 mM PBS was significantly different from those in 10 or 25 mM PBS.Similarly, ANOVA shows that ζ-potential values for both S. cerevisiae and P. fluorescens EVs were affected by changes in ionic strength.Tukey's HSD shows that the ζ-potential for P. fluorescens EVs at 1 mM PBS is significantly different than at 10 or 25 mM PBS, while the ζ-potential at all ionic strengths for S. cerevisiae EVs are significantly different.ANOVA also indicated that both S. cerevisiae and P. fluorescens EVs were affected by changes in HA concentration, regarding C/C 0 values.Tukey's HSD test indicated that the plateau for P. fluorescens EVs in 10 mg/L HA was significantly different from those in 0 or 1 mg/L HA.For S. cerevisiae EVs, Tukey's HSD only indicated that the C/C 0 value for the 0 mg/L solution of HA was different from the 10 mg/L sample.For ζ-potential values, ANOVA indicated that all EV types were affected by HA concentration, and Tukey's HSD showed that the ζ-potential for all EV types at 0 mg/L HA was significantly different from the ζpotential with 10 mg/L HA.
−46 From Figure 1, we noted that none of the EV populations deposited on the collector particles to a high extent at low ionic strength (1 mM).As ionic strength increased, different trends emerged for the different EV populations.For S. cerevisiae EVs, as ionic strength increased to 10 or 25 mM, their deposition to collector particles,and thus their attachment efficiency increased significantly, based upon both the standard error and Tukey's HSD test for C/C 0 values.A substantially lower C/C 0 value was also observed for S. aureus EVs at the highest ionic strength (25 mM), though this difference was not significantly different according to ANOVA.These results were corroborated with the ζ-potential data: as ionic strength increased, the surface potential of S. cerevisiae and S. aureus

Environmental Science & Technology
EVs became less negative, implying that the electrostatic repulsions between particles and collectors would be reduced to allow them to deposit to a greater extent.This observation is predicted by DLVO theory, and it follows a similar trend described in previous research regarding the effect of ionic strength on particle deposition, both for anthropogenic particles 36,42−44 and other biocolloids. 26,45,46A similar destabilization could be seen in the less negative ζ-potential for P. fluorescens EVs at higher ionic strengths, but this effect was not observed for the deposition on collectors, as represented by the similar C/C 0 values.This implies that there may be other factors influencing the deposition of P. fluorescens EVs, such as steric effects from other extracellular materials, hydrophobic effects, or unique interactions from the presence of lipopolysaccharides (LPS).−49 The external-facing parts of LPS consist of long chains of sugars (O-antigen) that could cause steric stabilization interactions with other EVs.Additionally, O-antigen is known to shield the negative charges of the lipid surface, 11 making P. fluorescens EVs less sensitive to changes in ionic strength.Such shielding would not be the case for Gram-positive or yeast EVs which lack LPS.Next, we considered the influence of HA in our column studies.2][43][44]50 We observed that the presence of the highest tested concentration of HA (10 mg/L) was associated with a greater level of breakthough for S. cerevisiae EVs (Figure 1). Thi trend is consistent with the ζpotential data.For the highest concentration of HA, the ζpotential of EVs at neutral pH became more negative (Table 1), which would imply greater electrostatic repulsion between particles.This is reflected in the reduced deposition on the collectors and thus a reduced attachment efficiency too.This result also aligns with existing research that shows the stabilizing effect of HA on colloidal particles.36,[42][43][44]50 In addition, previous work shows that lower concentrations (<2 mg/L) of HA have a less pronounced effect on colloidal deposition, 50 as also reported in this study.
By contrast, deposition of the P. fluorescens EVs increased in the presence of HA resulting in a lower C/C 0 value, and there was no significant effect on C/C 0 for S. aureus EVs with respect to HA concentrations.We recall that for all three organisms, this higher concentration of HA resulted in a more negative ζpotential at neutral pH (Table 1); thus, the effects of HA on EV stability cannot be explained entirely based on charge stabilization.That the particle deposition of P. fluorescens EVs increases with increasing HA concentration was particularly surprising: according to the ζ-potential data, a 10 mg/L addition of HA caused the ζ-potential to be more negative relative to the P. fluorescens EVs without HA, meaning that some factor other than electrostatic forces must be influencing the deposition of EVs.LPS has been shown to play a role as adhesins in biofilm formation. 51In addition, LPS has been shown to interact with HA, 52 which collectively may be the cause of this increased attachment.This departure from classic DLVO theory demonstrates that the transport of EVs must be evaluated with more than surface potential measurements.Unlike the attachment efficiencies of P. fluorescens EVs, the attachment efficiencies of S. aureus EVs are not significantly different.The lack of LPS on the surface of Gram-positive bacterial EVs could be the reason for this difference between bacterial species.
To make comparisons to other colloidal systems for which attachment efficiencies are known and to parameterize the transport model, α was calculated using eq 2, where the C/C 0 values used were those recorded at 2.25 bed volumes.Tables 2 and 3 present those calculated values.Other parameter values used in eq 2 to calculate α can be found in Table S1.
The α values for the three types of EVs we evaluated in this study are comparable to those reported for other biocolloids.For instance, for bacteriophage PRD1, values of α were reported from 0.0011 to 0.0053 at pH 7 for an ionic strength from 1 to 20 mM. 53These α values are on the same order of magnitude as those for EVs from S. cerevisiae and S. aureus, but they are still higher than those for P. fluorescens, again highlighting the possible role that the distinct properties of EVs derive from different parent species.
Minimal data exist for how the attachment efficiency of biocolloids varies as a function of the organic content in the surrounding media.One study examined the deposition of Cryptosporidium parvum oocysts which showed a reduced α from 0.87 to 0.18 when the organic content was present. 54nother similar study of C. parvum oocysts revealed that with the addition of 5 mg/L natural organic matter, the value of α dropped from 0.84 to 0.22, implying increased stability. 55These trends align with our data for S. cerevisiae EVs but not for either of the bacteria-derived EVs.
3.3.EV Removal as a Function of Darcy Velocity.The closed form solutions for collector efficiency 17,31 predict that collector efficiency for a particle of a given size and attachment efficiency should decrease with increasing velocity through the porous medium.Using S. cerevisiae EVs, column studies were carried out under a range of superficial flow rates (0.8, 0.4, and 0.2 mL/min) under the assumption that EV attachment efficiency and flow rates are independent of each other.The breakthrough curves for these systems are shown in Figure 2.These flow rates were chosen because this range fits into a typical groundwater flow rate range 56 and align with previous column experiment flow rates. 37,57Given a filter area of 7.85 × 10 −5 m, the corresponding Darcy velocities of these experiments were 1.7 × 10 −4 , 8.5 × 10 −5 , and 4.2 × 10 −5 m/s, respectively.According to eq 2, collector efficiency should decrease with increasing velocity through the porous medium.
Data obtained from experiments were then compared with calculations of anticipated concentrations using eq 2 using an average value of α for S. cerevisiae of 0.00496, determined from experiments in the absence of HA and at 1 mM ionic strength and pH 7 (Table 2).All other experimental conditions and model parameters were held constant for the purpose of comparison, changing only the value of the Darcy velocity.Figure 3 shows the results of calculations for the predicted EV concentration profile through the column for each of the flow rates.These calculations indicate that the effluent concentration from the column (normalized to influent concentration) is predicted to drop by approximately 15% as the flow rate decreases from 0.8 to 0.2 mL/min.However, we note that our observed EV effluent concentrations only dropped by approximately 3% (Table 2).
Table 4 compares the experimentally observed removal of EVs with the theoretically determined (model) removal.In general, the theoretical and observed removals were similar, especially in the case of 0.8 and 0.4 mL/min, and the observed trends align with the theoretical expectation that at a lower flow rate, there would be greater removal.However, in the case of the 0.2 mL/min flow rate, a divergence between the observed and expected values is statistically significant.It is noted that the differences in the observed C/C 0 values are not statistically significant (Table S8).
Consistent with filtration theory and previous experimental results, 58−60 the data show that as flow rate decreases, the EV deposition increases.While qualitative trends predicted by theory were observed, a statistically significant difference between predicted and observed differences in particle removal was observed at the lowest Darcy velocity.EV attachment efficiency should not be affected by a changing velocity but the collector efficiency will be.Furthermore, EV size and density are the only variable parameters unique to determining the collector efficiency; hence, estimates for these parameters may need to be better characterized for future iterations of this transport model.
The column experiments and corresponding model presented above may be used as the basis for making rough estimates of the spatial zone of influence that a given source of EVs might exert.This zone of influence would be relevant in predicting, for instance, the distance that a pathogenic organism could affect a target organism, or the distance over which genetic material from one organism might be transported in the environment.Such zones of influence might be considered in reverseengineering EVs to create delivery vehicles for agricultural supplements, using, for example, EVs as nutrient "taxis" to root systems.

ESTIMATIONS OF EV TRANSPORT IN POROUS MEDIA
By performing sensitivity analyses with this model, initial predictions for long-range transport of EVs can be determined and the model's limitations can be identified.Figure 4 shows the model output where the fraction of EVs removed are shown for  4.

Environmental Science & Technology
various Darcy velocities and values of α as a function of media depth (i.e., depth of packed, porous material).Figure 4a shows that for a bed depth equal to that of the column, less than 2% of the EVs are removed at a Darcy velocity of 10 −3 m/s, while at two orders of magnitude greater velocity, there is greater than 25% removal.To contextualize this observation, groundwater Darcy velocities range from 10 −3 to 10 −6 m/s. 56In addition, for α = 0.1, the model shows that the EV concentration is diminished to 10% of the initial concentration after moving through 0.15 m of saturated porous media, while for an α = 0.001, EV concentration does not drop to 10% until passing through almost 17 m of saturated porous media.Given the interest in some research sectors to think more specifically about engineering EVs to act as shuttles of nutrients 61 or even specific/desired genetic material, Figure 5a provides a sensitivity analysis for how the deposition of EVs might change in response to alterations to attachment efficiency, a parameter that engineers may be able change through modifications to EV surface properties.Figure 5b shows a sensitivity analysis for changes in model responses to changes in flow rate.While similar calculations for theoretical transport could be applied to other environmentally relevant nanomaterials, to our knowledge, this report contains the first application to EVs.Note that other model parameters were left unchanged in these analyses.
These sensitivity analyses reveal that under both conditional changes, there is a great potential to fine-tune EV transport through columns of saturated porous media.As seen in Figure 5a, even for minuscule increases in attachment efficiency, EVcollector deposition could be discouraged and EV transport increased.The same could be said of flow rate, where the natural changes to rainwater flow, for example, could dramatically alter the expected transport of EVs.
The transport of EVs depends on their environmental conditions such as solution chemistry.DLVO theory can help us to predict this transport, but some of the trends observed in this work deviated from theoretical expectations.The transport of EVs from the Gram-negative bacteria P. fluorescens seemed to consistently differ from expectations based on surface potential measurements, with many of the unique behaviors possibly stemming from the presence of LPS on the surface of their EVs.Continuing to expand the study of organisms' EVs beyond the three species evaluated here will help to make broader claims about EV transport.In addition, the effects of upstream isolation methods on downstream surface property analysis are necessary for data comparisons between EV studies.For instance, unpurified EV samples may be more accurately termed "extracellular biocolloidal particles" to take into account the mixture of colloidal particles produced by organisms, akin to the way humic substances are described.Overall, this work provides a foundation upon which the transport capabilities of EVs can be built.α values were determined for three microbial EVs, and a model was developed from this data.Through a sensitivity analysis, the model is shown to be sensitive to both flow velocity and α, allowing for predictions of the transport distance in porous media with respect to these two parameters.In particular, the sensitivity of the model to changes in α can be utilized by researchers to identify what surface properties are needed to transport EVs to a target distance in the environment.

a
Error is standard deviation for both measurement methods.b The Z-average diameter is the intensity-weighted, harmonic mean size.DLS error is based on three biological replicates (EVs were isolated from three independent cultures for each organism).c All ζ-potential values are reported at pH = 7.0 ± 0.2.ζ-potential values were compared for EVs derived from the same organism in the indicated ionic strengths and HA concentrations.Distinct letters indicate statistically different values.Statistical differences between different organisms' EVs were not evaluated.

Figure 1 .
Figure 1.Breakthrough curves for three EVs as a function of PBS ionic strength (left panels) and HA concentration (right panels).Breakthrough curves are measured at pH 7. The error bars are standard error for duplicate measurements.The indicated ionic strength (in mM) is the ionic strength of PBS; the indicated mass concentration (in mg/L) is the concentration of HA.Data graphed on the left column were all with 0 mg/L HA; data graphed on the right were all measured using 10 mM PBS conditions.

Figure 2 .
Figure 2. Breakthrough curves for S. cerevisiae EVs at three different flow rates in 1 mM PBS at pH 7. The error bars are standard error for duplicate measurements.

Figure 3 .
Figure 3. Model-determined normalized concentration profile of S. cerevisiae EVs through the column as a function of flow rate in 1 mM PBS at pH 7 with no HA.Model-predicted values of the outlet concentration are listed in Table4.

Figure 4 .
Figure 4. Model prediction for EV transport through saturated porous media.(a) Projected concentration profile at a distance of 0.15 m as a function Darcy velocity.(b) Projected concentration profile for a Darcy velocity of 10 −4 m/s, noting when C/C 0 = 0.1, as a function of α.

Figure 5 .
Figure 5. (a) Model sensitivity to changes in α using a Darcy velocity of 5.4 × 10 −5 m/s.(b) Model sensitivity to changes in flow rate with an α = 0.00496.

. Vesicle Isolation and Sample Preparation. EVs
were isolated with modifications to the protocol in McMillan et al. and Rodriguez and Kuehn.

Table 1 .
EV Size Ranges and ζ-Potential Values

Table 2 .
Calculated Values of α for Three Populations of EVs in Various Ionic Strengths Error is standard error based on duplicate measurements.b C/C 0 values were compared for EVs derived from the same organism.Distinct letters indicate statistically different values.Statistical differences between different organisms' EVs were not evaluated. a

Table 3 .
Calculated Values of α for Three Populations of EVs in Various HA Concentrations a Error is standard error based on duplicate measurements.b C/C 0 values were compared for EVs derived from the same organism.Distinct letters indicate statistically different values.Statistical differences between different organisms' EVs were not evaluated.

Table 4 .
Theoretical and Observed Values of EV Removal a a Error is standard error based on duplicate measurements.