Investigating and modeling the regulation of extracellular antibiotic resistance gene bioavailability by naturally occurring nanoparticles

Abstract

Extracellular antibiotic resistance genes (eARGs) are widespread in the environment and can genetically transform bacteria. This work examined the role of environmentally relevant nanoparticles (NPs) in regulating eARG bioavailability. eARGs extracted from antibiotic-resistant B. subtilis were incubated with non-resistant recipient B. subtilis cells. In the mixture, particle type (either humic acid coated nanoparticles (HASNPs) or their micron-sized counterpart (HASPs)), DNase I concentration, and eARG type were systematically varied. Transformants were counted on selective media. Particles decreased bacterial growth and eARG bioavailability in sterile systems. When DNase I was present (≥5 μg/mL), particles increased transformation via chromosomal (but not plasmid-borne) eARGs. HASNPs increased transformation more than HASPs, indicating that the nanoscale increases eARG bioavailability. These results were also modeled via particle aggregation theory, which represented eARG-bacteria interactions as transport leading to collision, followed by attachment. Using attachment efficiency as a fitting factor, the model predicted transformant concentrations within 35% of experimental data. These results confirm the ability of NPs to increase eARG bioavailability and suggest that particle aggregation theory may be a simplified and suitable framework to broadly predict eARG uptake.

Graphical Abstract

1. Introduction

The global rise in antibiotic resistance has led to worldwide increases in mortality, morbidity, and economic costs.^1–3 Mobile genetic elements, including extracellular antibiotic resistance genes (eARGs), may be significant spreaders. eARGs have been quantified in freshwater⁴, sediment⁵, animal feed operations^{3, 6}, drinking water treatment plants^{3, 7} and wastewater treatment plants⁸, among other environments. The environment contains naturally competent bacterial species that may ingest eARGs and incorporate them into their genomes via horizontal gene transfer in a process called natural transformation.

Natural transformation is a significant ecological phenomenon which contributes to an extensive microbial gene network. This environmental process has been detected in situ^{9, 10}, and characterized on the bench-scale.¹¹ The significance of this process is notably enhanced by the interaction of extracellular genes with environmental components. In nuclease-containing systems, extracellular DNA (eDNA) bound to natural substances such as montmorillonite ¹², illite ¹³, humic acid ¹⁴, and sand ¹⁵ are able to transform bacteria at greater rates than free eDNA. Because nucleases are ubiquitous in the environment, it is reasonable to consider particle-association as a driver for natural transformation. Among the particles that may associate eDNA are naturally ubiquitous nanoparticles (NPs) which have been shown to irreversibly bind eARGs¹⁶ and protect them from enzymatic degradation.¹⁷ Based on these findings, this work hypothesizes that NPs may also mediate natural transformation. The ability of NPs, compared to larger particles, is important to study because nano-emergent traits, such as increased surface area to volume ratios, may facilitate interaction with biological components. NPs also tend to be more mobile¹⁸ and bioavailable¹⁹ than their larger counterparts, and may enhance the spread of bound eARGs in the environment.

To characterize the role of NPs in regulating eARG bioavailability, this work reports on: 1) the measurement of natural transformation rates of bacteria via eARGs sorbed to particles of various sizes (nano- and micron-scale) and 2) models formulated to describe transformation rates under a broad range of conditions. Culture-based methods are used to quantify transformation of naturally competent B. subtilis by various eARGs pre-equilibrated with nucleases and humic acid coated silica particles. It is hypothesized that the nanoscale silica particle would be most effective at facilitating transformation because it has greater total surface for gene or bacteria attachment. The experimental transformation rates are used to construct a model describing natural transformation as an aggregation process between eARGs (free and sorbed) and bacteria. Particle aggregation theory, initially theorized by Smulochowski²⁰, has been used to inform models of various environmental processes such as phage-host attachment.²¹ It is hypothesized here that particle-particle interactions described by Smulochowski’s theory may realistically depict the attachment of eDNA to bacterial surfaces. This is the first attempt not only to assess the ability of NPs to mediate natural transformation, but to model natural transformation via particle-associated eARGs. This effort is made to better understand the environmental rise of antibiotic resistance.

2. Methods

2.1. Bacterial strains

All bacteria used in experimentation were Bacillus subtilis 168 from the Bacillus Genetic Stock Center (BGSC) (Ohio State University). Strains used include B. subtilis 1A1 (non-resistant), B. subtilis 1A189 (multi-drug, including acriflavine, resistant), B. subtilis 1A354 (sulfanilamide resistant) and B. subtilis 1A491 (trimethoprim resistant).

B. subtilis 1A189 includes a point mutation (acfA, which is an Adenine-Thymine base-pair deletion) in its genomic DNA within the promoter region of the 1203 bp blt gene. The blt gene is part of the bltR-blt-bltD genome segment. The full sequence, location of each gene and location of mutation are given in SI. The mutant genome segment codes for efflux-mediated resistance to multiple antibiotics including acriflavine.¹

B. subtilis 1A354 may include multiple plasmid-bound sulfanilamide resistance genes, such as sul1, sul2 or sul3. These sul genes code to produce an alternative dihydropteroate synthetase with low sulfanilamide affinity.²² B. subtilis 1A491 includes multiple mutations, likely substitutions within the dfrA24 gene^{23, 24}, and is trimethoprim resistant by way of modifying the dihydrofolate reductase.²⁴ These strains were chosen so that their mutations could be compared. All strains were revived and cultured on nutrient agar at 37°C according to BGSC instructions. Strains used and their properties are summarized in SI.

2.2. DNA extraction and PCR amplification

Genomic DNA from each antibiotic resistant B. subtilis strain was extracted using the DNeasy UltraClean Microbial Kit from Qiagen (Hilden, Germany), according to the manufacturer’s instructions. Extraction was confirmed by quantification of DNA using the Qubit 2.0 fluorimeter.

Fragments of the gene blt from were amplified from extracted B. subtilis 1A189 DNA using the polymerase chain reaction (PCR). The PCR protocol is given in SI. Five fragments of various sizes (191 bp, 400 bp, 906 bp, 1100 bp and 1500 bp) were extracted. Each fragment was chosen, using widely available B. subtilis 168 sequencing data, such that the point mutation acfA is symmetrically flanked. The locations and sizes of fragments, along with primers, are demonstrated in SI. These fragments were prepared to test for the impact of flanking region length on transformation. The 906 bp fragment was also amplified as single stranded DNA (ssDNA) to test the impact of strand conformation.

2.3. Particle preparation

Two model colloids were used in transformation studies: 100 nm silica nanoparticles (Nanocomposix, San Diego, CA) surface functionalized with humic acid (Sigma, St. Louis, MO) (HASNPs) and 1 μm silica nanoparticles (Nanocomposix, San Diego, CA) surface functionalized with humic acid (HASPs). HASNPs were used as model NPs because the functionalization of environmentally relevant silica NPs with humic acid allowed for the adsorption of eDNA to particles.¹⁶ HASPs were used as the micron-sized equivalent for comparison.

HASNPs and HASPs were prepared as described in SI. They were characterized for their hydrodynamic diameter and calculated zeta potentials by dynamic light scattering (DLS). These results are given in SI.

2.4. DNA adsorption and DNase I exposure

eARGs were equilibrated with particles and DNase I prior to natural transformation assays. DNase I was added so that the impact of particle-mediated protection of genes¹⁷ could be assessed on transformation. The DNA adsorption and DNase I exposure procedure is described in SI. Extracted B. subtilis 1A189 DNA was exposed to various DNase I concentrations (0, 0.5, 1, 5, 10, 20 μg/mL) to determine the extent to which the protective effect of particles on eDNA impacts transformation. The remaining eARGs (the 5 PCR amplified blt fragment, the single stranded 906 bp blt fragment and DNA extracted from B. subtilis 1A354 and B. subtilis 1A491) were exposed to a single DNase I concentration (1 μg/mL). For each eARG type, the adsorption and exposure process were performed for each particle condition (HASNP, HASP or particle-free) in triplicate.

2.5. Natural transformation assay

Transformation tests were performed with non-antibiotic resistant B. subtilis 1A1 as the recipient cell. Transforming eDNA was extracted from B. subtilis 1A189 for tests with DNase I variations, PCR amplified fragments of B. subtilis 1A189 for tests with variations in fragment size and strand conformation, B. subtilis 1A354 for testing sulfanilamide resistance and B. 1A491 for testing trimethoprim resistance. The full transformation assay was adapted from He et al. (2019)¹ and is described in SI. A schematic summarizing the full transformation procedure is given in Figure 1.

Figure 1:

Schematic of natural transformation assay in which eARGs extracted from donor bacteria are amplified by PCR and used to transform nonresistant recipient bacterial cells.

2.6. Model Simulations

2.6.1. Model formulation

Natural transformation of particle associated eARGs is modeled through three reactions.

$$\mathrm{N}+\mathrm{G}\leftrightarrow \mathrm{L}$$ (Reaction 1)

$$\mathrm{B}+\mathrm{L}\stackrel{{\mathrm{\gamma }\mathrm{k}}_2}{\to }\mathrm{T}$$ (Reaction 2)

$$\mathrm{B}+\mathrm{G}\stackrel{{\mathrm{\gamma }\mathrm{k}}_3}{\to }\mathrm{T}$$ (Reaction 3)

N, G, L, B, and T represent the mass concentration of nanoparticles, free eARGs, sorbed eARGs untransformed bacterial cells, and transformed bacterial cells, respectively. Reaction 1 describes a reversible adsorption at equilibrium between eARGs and particles. Reaction 2 and 3 represent the uptake of free and dissolved genes, respectively, by bacteria as a non-reversible “aggregation” between DNA and cells.

The rate constants k₂ and k₃ consist of two parameters that encompass a modified Smoluchowski approach to tracking particle aggregation: α (attachment efficiency) and β (collision frequency). β (with units of inverse particle concentration per time), represents the frequency of particle collisions via three pathways: Brownian motion (BR), differential settling (DS), and shearing (S). Here, the rectilinear model²⁰ is utilized to calculate β. The effect of compressing streamlines during particle approach is therefore not considered. For spherical particles this hydrodynamic retardation has the effect of significantly overestimating collisions between particles of very different sizes. However, given the non-spherical nature of the particles involved and the assumptions applied, the use of the curvilinear model for the collision kernel was not considered warranted. The collision frequencies from the transport pathways are additive. The full analytical equations for β are provided in SI.

Attachment upon collision between eARGs and bacteria is a function of several factors - such as the binding sites available on the cell surface²⁵, the presence of protein-like binding factors ²⁶, pH ²⁷ and sequence homology.^{10, 28} The probability that a particle or gene will attach to a bacterium is depicted in this model by α. Theory for calculating α is highly restrictive, therefore it is treated as a fitting factor that corrects for the assumption of rectilinear particle trajectories. Furthermore, whether or not an attached gene is integrated into the chromosome is dependent on various factors – such as divalent cation concentration²⁹, the presence of competence proteins ³⁰ and sequence homology.²⁷ Therefore, the additional correction factor γ is used to convert αβ to the number of attachments that subsequently lead to transformed bacteria.

With these factors in mind, along with assumptions of no cell growth or decay, the following population balances (based on Reactions 1–3) and expressions for k2 and k3 can be made:

$$\frac{\mathrm{d}\mathrm{B}}{\mathrm{d}\mathrm{t}}=-{\mathrm{\gamma }(\mathrm{k}}_2\mathrm{L}\mathrm{B}+{\mathrm{k}}_3\mathrm{G}\mathrm{B})$$ (Equation 1)

$$\frac{\mathrm{d}\mathrm{T}}{\mathrm{d}\mathrm{t}}=\mathrm{\gamma }({\mathrm{k}}_2\mathrm{L}\mathrm{B}+{\mathrm{k}}_3\mathrm{G}\mathrm{B})$$ (Equation 2)

$${\mathrm{k}}_2=\frac{{\mathrm{\alpha }}_{\mathrm{L}\mathrm{B}}{\mathrm{\beta }}_{\mathrm{L}\mathrm{B}}}{\gamma {\rho }_L{V}_L}\left[=\right]\frac{L^3}{M*T}$$ (Equation 3)

$${\mathrm{k}}_3=\frac{{\mathrm{\alpha }}_{\mathrm{G}\mathrm{B}}{\mathrm{\beta }}_{\mathrm{G}\mathrm{B}}}{\gamma {\rho }_G{V}_G}\left[=\right]\frac{L^3}{M*T}$$ (Equation 4)

where k₂ describes the aggregation between attached genes and bacteria and k₃ describes the aggregation between dissolved genes and bacteria. A full derivation of these equations and parameter enumeration is available in the SI.

Reaction 1 is assumed to occur rapidly such that the fraction of sorbed eARGs, f_p, can estimated using partition coefficients, P, calculated as the slope of the linear portion of the pertinent sorption isotherms.

$${\mathrm{f}}_{\mathrm{p}}=\frac{\mathrm{P}\mathrm{N}}{1+\mathrm{N}}$$ (Equation 5)

The model also seeks to represent the loss of genetic material due to enzymatic degradation. This is assumed to occur as a pseudo-first order reaction with degradation rate constants k_d,1 and k_d,2, which describe degradation kinetics of sorbed and free eARGs, respectively. These rate constants can be determined from kinetic eARG degradation data. A mass balance can account for the total amount of genetic material, or c_T.

$$\frac{\mathrm{d}{\mathrm{c}}_{\mathrm{T}}}{\mathrm{d}\mathrm{t}}=-{\mathrm{k}}_{\mathrm{d},1}{\mathrm{f}}_{\mathrm{p}}{\mathrm{c}}_{\mathrm{T}}-{\mathrm{k}}_{\mathrm{d},2}(1-{\mathrm{f}}_{\mathrm{p}}){\mathrm{c}}_{\mathrm{T}}$$ (Equation 6)

2.6.2. Calibration

Model parameters were estimated from literature and experimental data. The values and methods used to estimate them are summarized in SI. These values were used to convert transformant counts in CFU/ml to mass concentration. Calibration consisted of determining values for α that yielded a satisfactory fit to experimental data using B. subtilis 1A189 genomic DNA to transform B. subtilis 1A1, in a system with 1 μg/mL DNase I over 90 minutes. Calibrations were performed for both HASNP and particle-free datasets (Figure 3). α_GB was calibrated to match the experimental T determined in particle-free systems. Then, with α_GB held constant, the initial particle concentration N₀ was set to a value of 0.5 g/mL and α_LB was iterated until T matched experimental values at 90 minutes.

Figure 3:

Transformation frequencies (f) in systems with varying DNase I concentrations and either HASNPs, HASPs or no particles.

2.6.3. Validation

The model was validated using transformation assays in HASP systems (using B. subtilis 1A189 genomic DNA to transform B. subtilis 1A1 in a system with 1 μg/mL DNase I over 90 minutes) (Figure 3). All parameters were held constant except d_L, which was changed to of 2.246 μm for HASPs as determined by Zetasizer (Malvern Panalytical, Malvern, UK) measurements. Since it is reasonable to assume that less eARGs adsorb to the larger particle, which has smaller specific surface area, the partition coefficient determined for the NPs was adjusted for the micron-scale HASPs by scaling the specific surface area (see SI). The model-predicted T and the experimental T after 90 minutes were compared.

2.7. Analysis

All data points were collected in triplicate and standard error of the mean was determined. The significance of particle addition, resistance gene type and/or DNA properties on transformation rates was determined by analysis of variance (ANOVA). Significance was assumed at p < 0.05.

Model parameters (summarized in SI) were used to determine collision frequency parameters and rate constants. The mass and population balances in Equations 1, 2 and 6 were solved numerically using Euler integration using MAT-LAB R2019a.

3. Results and discussion

3.1. Transformation assays

3.1.1. Particle effects on bacterial concentration

With the addition of either HASNPs (p<0.041) or HASPs (p<0.05), the total concentration of bacteria, B, decreased significantly (Figure 2). This implies that particles inhibited the growth of bacteria. In some cases, decreased bacterial growth has been observed in bacterial cells associated with particle surfaces due to less cell surface available for substrate uptake ^{31, 32}, higher maintenance coefficient ³¹, or substrate transport limitations.³³ Such phenomena may be contributing to decreased B. subtilis growth in association with HASPs and HASNPs.

Figure 2:

Bacterial counts, B(CFU/mL), observed in negative control tests for the transformation protocol (No eARGs or DNase I) with various particle types.

3.1.2. Particle effects on transformation in systems with DNase I

With no DNase I present, transformation frequencies (f) are highest for particle-free systems (Figure 3). A significant decrease in f is observed with the addition of HASPs (p< 0.000768) or HASNPs (p<0.000423). This implies that particles may hinder transformation in sterile systems. This finding is in accordance with previous studies^{12–14, 34}, which have suggested that a fraction of adsorbed DNA is available for bacterial uptake, while a fraction is not.^{13, 35} Sorption of eDNA to the particle surface may decrease the total fraction of bioavailable genes. This observed inhibitory effect on transformation is greater for HASPs than HASNPs (p<0.0247). In some cases, bacterial uptake is more efficient when bacteria is bound directly to the mineral surface.³⁴ If NPs have greater surface area per volume, they may also have more bacterial binding sites.

At 0.5 μg/mL DNase I concentrations, the inhibitory effect of particles is no longer observed (p<0.331). When DNase I concentrations are increased to 1 μg/mL or more, there are significant increases in f observed for both HASP (p<3e-6), and HASNP (p<3.38e-5) systems, as compared to particle-free systems. This implies particles may increase natural transformation frequencies when sufficient nuclease is present, possibly because particles protect eDNA from degradation prior to uptake.¹⁷ Further, transformation frequency was significantly greater in HASNP systems compared to HASP systems at 20 μg/mL DNase I concentrations (p<0.00113). Enzymatic degradation of eDNA may be inhibited by particle adsorption to eDNA or to DNase I. NPs have a greater surface area per volume and potentially a greater number of binding sites that may allow eARGs or nucleases to bind.

3.1.3. Effect of eARG fragment size and strand conformation

No significant effect of fragment size (in eARG fragments with a symmetrically flanked point mutation) was observed in the size range tested, in systems with HASNPs (p<0.147), HASPs (p<0.766) or no particles (p<0.796) (Figure 4A). Larger flanking regions have been shown to aid transformation by facilitating homologous recombination.³⁶ Studies that have demonstrated such correlations have used donor fragments that range up to hundreds of thousands of base pairs long.³⁷ Perhaps, the variation in flanking region size tested here was not large enough to show significant differences in transformation frequency. Also, none of the PCR-amplified fragments tested included the full blt gene. Differences in transformation may have been more easily discernable with the full gene, as bacterial populations are more likely to retain a gene if taken up in entirety.³⁸

Figure 4:

Transformation frequencies (f) upon exposure to eARG fragments of (A) various sizes and (B) various strand conformations and (C) various mutated eARGs in systems with HASNPs, HASPs or no particles. * Denotes statistically significant difference between groups determined by a p value <0.05.

No significant differences in transformation frequencies via double-stranded (dsDNA) and single-stranded DNA (ssDNA) was observed for systems with HASNPs (p<0.679), HASPs (p<0.102) or no particles (p<0.419) (Figure 4B). Previous studies have noted an overall tendency for dsDNA to transform B. subtilis more effectively than ssDNA at neutral pH, but reported transformant counts were about 1–2 orders of magnitude greater for dsDNA than ssDNA.³⁹ This is comparable to differences observed here in particle-free systems. No significant difference in transformation frequency was detected between HASP and HASNP systems for dsDNA (p<0.162) or ssDNA (p<0.0686).

3.1.4. Particle-associated transformation via various eARGs

Significant differences in transformation frequency between the three genes were observed for HASNP (p<0.000178), HASP (3.6e-6) and particle-free systems (p<0.000811). This implies that transformation rates vary greatly between eARGs with different mutations, and that particle-effects have comparatively less influence. In particle-free systems, sul genes have the highest f values, followed by blt and then by dfrA24. dfrA24 is a chromosomal gene with multiple mutations, whereas blt, also a chromosomal gene, has a single mutation. It is possible that recipient cells require less metabolic energy to express plasmid-encoded sul, which does not require chromosomal integration for expression¹⁰, as opposed to chromosomal eARGs. Further, incorporating multiple mutations into the genome requires larger segments of DNA to be recombined into the recipient cell chromosome, and therefore greater energy cost and sequence homology. This may explain why frequencies are higher for blt as compared to dfrA24

Transformation frequencies significantly increase with the addition of HASNPs (p<7.69e-6 for blt, p<0.0245 for dfrA24) or HASPs (p<3.05e-10 for blt, p<0.0215 for dfrA24), with no significant difference observed between them (p<0.0877 for blt, p<0.506 for dfrA24). This demonstrates the ability of particles, including NPs, to improve transformability of multiple types of environmentally relevant chromosomal eARGs.

Transformations via sul, on the other hand, are not significantly impacted by the addition of either HASNPs (p<0.162) or HASPs (p<0.0553). In previous studies, plasmids in soil have demonstrated lower transforming activity than chromosomal genes, even when they remain physically stable.⁴⁰ Some part of particle association may inhibit the transformability of plasmid DNA.

3.5. Particle aggregation modeling

3.5.1. Model predictions

The calibration of the model using α as a fitting parameter yielded α values (SI) on the order of magnitude of those documented in previous experiments.⁴¹ Therefore, the calibration step was considered to provide reasonable corrections.

During validation with the HASP dataset, the model predicted a T value within 35% of the experimental result (Figure 6). The T measured in HASP systems was 0.0153 kg/m³, whereas the model predicted a T of ~ 0.010 kg/m³. Considering the simplifications applied to the model, the difference between the predicted and experimental T was acceptable for the purposes of validation, confirming the ability of particle aggregation theory to closely predict bacterial transformation rates.

Figure 6:

The contribution of Brownian motion (BR), differential settling (DS) and shear forces (S) on collision frequencies (β) between bacteria and (A) particle-bound eARGs and (B) free eARGs.

The model has unique value in its capacity to assess particle-level interactions that may govern the transfer of genes on particles. Possible contributions of various collision mechanisms (β) leading to the association of genes and bacteria are shown in Figure 7 below.

Figure 7:

Sensitivity of model to (A) eARG-particle complex diameter, d_L (B) attachment efficiency between particle bound eARG and bacteria, α_LB, and (C) partition coefficient, P, during adsorption.

Collisions between bacteria and free or NP-bound eARGs are predicted to be dominated by Brownian motion (Figure 7). For HASP-bound genes, shear forces increase greatly and override Brownian motion as the most dominant force, as expected. In the rectilinear model (at slow mixing and ambient temperature), Brownian motion is the predictably dominant mechanism of collision when particles are <1 μm while shear dominates for larger particles.⁴² The model also estimates that total collisions increase when particles are added and are greatest for the micron-sized particle compared to the NP. However, the rectilinear model is known to overpredict collision frequencies, particularly for larger particles.⁴²

3.5.1. Model limitations

To estimate adsorption parameters for HASPs where data was unavailable, the model assumes that adsorption is proportional to particle surface area. In reality, adsorption may be dependent on various factors. A uniform particle size for each particle type is also assumed, even though particles may have varying sizes due to heterogenous humic acid sorption on the particle surface. Additionally, the attachment efficiency parameter α_LB was calibrated for HASNP systems but is applied to HASP systems. In reality, α may vary based on particle size, surface chemistry, sorption capacity of eARGs, and more. Variations in α may also be due to assumptions inherent to the rectilinear model, which tends to overpredict collisions for larger particles²⁰ and this would be reflected in the α calibration. There were also simplifications to the biology of the system – such as the assumptions of no growth or decay, consistent γ, lack of consideration of gene properties, cellular competence, DNA conformation and more. As previously stated, the assumption of spherical particles following rectilinear collision trajectories may also distort the calculated interactions between particulate objects that are inherently non-spherical.

3.5.2. Sensitivity analysis

Though complex biological factors cannot easily be incorporated into the model, the contribution of assumptions about particle size, attachment efficiency and partition coefficient to model underpredictions were assessed. The values of d_L, α_LB and P were iterated to determine the extent to which these parameters caused variations between model-predicted and experimental T values for HASP systems.

Considering Figure 7A, T at 90 minutes can be reached with a d_L of less than 180 nm. However, this is not a realistic diameter for HASPs, because the silica nanosphere core is manufactured to be 1 μm in diameter. Therefore, the possible variation between measured and actual diameter of particles does not entirely explain discrepancies between model and experimental T predictions. T is also not particularly sensitive to α_LB (Figure 7B), suggesting that assumptions inherent to α do not particularly influence the effectiveness of the model. It can be concluded, then, that the rectilinear model may be a suitable framework to predict particle-mediated gene transfer processes since collision overpredictions influencing α_LB would not significantly alter the model outcome. Of the parameters considered, the model is most sensitive to P (Figure 7C). It is evident that without changes to d_L, a P value between 0 and 0.1 could lead to a prediction of T equal to the experimental value (0.0153 kg/m³).

The sensitivity of the model to d_L and P implies that the assumptions used to calculate these parameters are likely major contributors to deviations between model-predicted and experimental values. In experiments, the size of the particle does not significantly impact T at 1 μg/mL DNase I (Figure 3), but the model predicts a large difference in T based on particle size. The model, therefore, overpredicts the role of particle surface area. This may also be due to the model’s inability to account for other simultaneous processes, such as eARG conformation changes.

3.5.3. Environmental implications

This work establishes, for the first time, the ability of naturally occurring NP-associated genetic material to confer resistance to bacterial populations via natural transformation. Based on these findings, it can be speculated that in real environmental systems where extracellular genetic material is prone to nuclease degradation (such as after a cell lysis event), nanoparticle association may enhance gene transfer between microbes– to a greater extent than larger particles. Nanoparticles have not yet been considered as vectors for antibiotic resistance spread. Understanding their role in enhancing eARG bioavailability may have major implications for the tracking and characterization of antibiotic resistance.

The ability to model this phenomenon is also a step in better understanding the role eARGs play in propagating resistance. The developed model was capable of broadly predicting transformation frequencies in various conditions. This capacity implies that classical particle aggregation theory is a useful tool for quantifying extracellular gene uptake by microbes. This approach is valuable because, though subject to limitations, it allows for a simplified calculation based on available data that can closely predict transformation rates. It provides the ability to test effects of simple modifications in gene or particle properties on bioavailability. Such a model may realistically be incorporated into larger and more comprehensive simulations that assess the spread of antibiotic resistance in the environment.

Figure 5:

Model-predicted concentration of transformed bacteria (T) as a function of time (t) for HASP systems (N₀=0.5 μg/mL). Star depicts experimental value of T at 90 minutes.