Dissecting the effects of antibiotics on horizontal gene transfer: Analysis suggests a critical role of selection dynamics

Abstract

Horizontal gene transfer (HGT) is a major mechanism responsible for the spread of antibiotic resistance. Conversely, it is often assumed that antibiotics promote HGT. Careful dissection of the literature, however, suggests a lack of conclusive evidence supporting this notion in general. This is largely due to the lack of well-defined quantitative experiments to address this question in an unambiguous manner. In this review, we discuss the extent to which HGT is responsible for the spread of antibiotic resistance and examine what is known about the effect of antibiotics on the HGT dynamics. We focus on conjugation, which is the dominant mode of HGT responsible for spreading antibiotic resistance on the global scale. Our analysis reveals a need to design experiments to quantify HGT in such a way to facilitate rigorous data interpretation. Such measurements are critical for developing novel strategies to combat resistance spread through HGT.

Introduction

Lateral or horizontal gene transfer (HGT) refers to the transfer of genetic material within or between species, separate from the vertical descent of genes from parent to offspring. Since its initial discovery in 1928 [1], HGT has increasingly been appreciated as a driving force in microbial evolution [2,3]. It is estimated that up to 17% of the Escherichia coli genome is derived from previous HGT events, and up to 25% of genomes of other bacterial species [4].

In bacteria, HGT occurs through one of three major mechanisms: conjugation, transformation and transduction [5]. In conjugation, DNA is transferred from donor to recipient through direct cell-to-cell contact [5–7]. Conjugation occurs either through plasmid transfer or transmission of chromosomally integrated conjugation elements (ICEs) [8], including conjugative transposons (CTns) [9]. Conjugative plasmids can replicate autonomously. They are self-transmissible when carrying both the genes required for transfer machinery, and the replication origin recognized by the transfer machinery. Conjugative plasmids that are not transmissible are called mobilizable; these plasmids do not encode the conjugation machinery, but they transferrable by the machinery encoded elsewhere [10]. Collectively, conjugative plasmids are estimated to account for about half of all plasmids [11]. Unlike plasmids, integrated conjugative elements (ICEs) reside in the host chromosome but can be excised and transferred by conjugation [12]. It is thought that conjugation is responsible for the majority of antibiotic resistance spread, since there is a high prevalence of plasmids, plasmids can have broad host ranges [13], and many plasmids carry an abundance of antibiotic resistance genes [14–16].

In transformation, naked DNA is taken up by bacteria from the environment and incorporated into the genome [17,18] while in transduction, gene transfer is mediated by bacteriophages [19]. Transduction and transformation can transfer a wide range of sequences including antibiotic resistance cassettes [20–24]. Transduction occurs following specific recognition between the phage and its cognate receptor on the bacterial surface. This specificity results in a limited host range for many phages and is therefore believed to have contributed less to the spread of antibiotic resistance compared to conjugation [11]. Natural transformation often occurs in a transient physiological state of competence induced by environmental cues such as nutrient access or cell density [25]. Although bioinformatic analyses reveal that the majority of bacteria possess genes homologous to known competence genes [26], it is unclear whether these bacteria undergo transformation [27,28]. Thus, studies that suggest few naturally transformable organisms exist (i.e.[29]) typically assume that transformation does not play a significant role in spreading of antibiotic resistance genes, though further investigation is warranted.

Horizontally transferred genes can encode diverse functions, including metabolic traits [9,30], virulence factors [31,32], and antibiotic resistance [33–35]. Mobile genetic elements (MGEs) carrying resistance are members of the communal gene pool accessible to diverse microbes [10]. Since HGT can cross phylogenetic boundaries, it can greatly accelerate the spread of resistance genes [36]. Indeed, it has been suggested that HGT has enabled the spread of resistance to most antibiotics commercially available [37–39]. Conversely, the use of antibiotics can potentially influence dynamics of HGT, and thus the spread of antibiotic resistance. However, an antibiotic can affect the efficiency of HGT at the single-cell level, modulate the overall HGT dynamics through selection, or both. Elucidation of these effects requires rigorous design of experiments to tease apart these different effects.

Conjugative spread of antibiotic resistance

HGT and antibiotic selection both play critical roles in shaping the global resistance landscape [40,41]. HGT specifically enables the local dispersal of resistance genes from a source species to other strains or species within a local context such as the gut of a patient [42,43]. Carriers of newly resistant bacteria (such as people, animals, or products) continuously move and interact. This facilitated dispersal enables dissemination of the newly resistant bacteria on a geographic scale. In new environments, resistant bacteria further propagate the resistance determinants through HGT. Antibiotic use can magnify the overall effect of HGT by enriching for the resistant progeny. Selection can increase the abundance of cells capable of HGT, making further HGT events more likely [44]. Together, this has likely enabled the global spread of resistance [45,46].

This is best exemplified by the spread of β-lactamase (Bla) variants. Bla enzymes confer resistance to β-lactams, which are one of the oldest and most widely prescribed antibiotic class [38]. Bla variants, including extended-spectrum β-lactamases (ESBLs [47], are often encoded on plasmids [48,49], and many bla genes likely spread via conjugation [50,51]. For example, MGE-associated CTX-M (a common bla variant) in Enterobacteriaecae species shares adjacent gene sequences to those from Kluyvera species [52], suggesting lateral acquisition. Similarly, the diversity of species carrying sequence-identical bla_NDM-1 (a metallo- β–lactamase conferring resistance to carbapenem drugs) on conjugative plasmids implicates the role of conjugation. Indeed, from analyses of sewage and tap water in New Delhi, one study found the presence of bla_NDM-1 in over 20 diverse strains, many of which were never before associated with ESBL resistance; every isolate could transfer its plasmid via conjugation under certain experimental conditions [53]. The intense β-lactam selection pressure over the past several decades has likely facilitated the prevalence of such examples.

Other common horizontally transferred resistance genes are similarly linked to antibiotic exposure. For example, vancomycin is typically considered a last-resort drug for methicillin-resistant Staphylococcus aureus (MRSA) infections. Although VanA vancomycin resistance is typically associated with the transposable element Tn1546 common to vancomycin-resistant Enterococcus (VRE)[54], VanA-expressing vancomycin-resistant Staphylococcus aureus (VRSA) has begun to emerge. Analysis of clinical samples shows that VRE indeed can transfer VanA-encoding Tn1546 to MRSA by conjugation. Similarly, despite fluoroquinolone (FQ) resistance initially driven by mutations that cause target modification [55] or increased efflux activity [56], researchers have recently identified a new form of FQ resistance mediated by plasmid-encoded Qnr protein. Qnr binds to DNA gyrase and topoisomerase IV [57,58], conferring moderate protection of these components against FQ [59]. Studies have since reported increasing prevalence of qnr and its variants on conjugative plasmids isolated from the clinic [60,61], often associated with multi-resistant phenotypes such as ESBL producers.

Our gut microbiota harbors an abundance of resistance genes [62,63], and can act as a reservoir by which other commensal or pathogenic bacteria can acquire resistance through HGT [64–66]. Antibiotics can provide a favorable environment for transfer by enriching for resistance already present at low levels [42,44]. Such selection can also influence the abundance of resistance to other antibiotics if the two (or more) genes are physically linked on the same MGE. For example, one study detected high levels of qnr-mediated FQ resistance in the gut flora of children that were treated with non-FQ antibiotics [67]. Different factors may drive resistance landscapes depending on the environment, however. One study concluded that bacterial community composition, not HGT, is the major determinant in shaping the resistance landscape in natural soil environments [68].

Given the prevalence of resistance spread through conjugation, it is tempting to assume that use of antibiotics promotes HGT rates. Undoubtedly, application of antibiotics in human medicine and animal husbandry has changed the ecological landscape of resistance [69], and has likely influenced HGT through modulating the abundance of species carrying resistance located on MGEs [70]. It is even speculated that antibiotics may modulate the tempo of HGT by selecting for cells with higher transfer efficiencies [45]. However, the spread of antibiotic resistance is not always correlated with antibiotic use. For example, ESBL genes had evolved and spread through conjugation millions of years before the modern practices of antibiotics [71,72]. In general, the extent to which antibiotics modulate the HGT rate is a nuanced question that depends on several factors. To elucidate this question, it is critical to clarify the definition of the HGT rate, and to properly design experiments to generate conclusive results.

Reaction kinetics of conjugation

At the population level, conjugation can be approximated as a biomolecular reaction between the donor (D) and the recipient (R), resulting in the transconjugant (T), where D, R, and T each represents the respective cell densities [73]. The rate of generation for T by the two parents can be written as:

$$\frac{\mathit{dT}}{\mathit{dt}}=\eta \mathit{RD}$$ (1)

where η is the rate constant for the interaction, which we term as conjugation efficiency.

During a short time window of conjugation (Δt), if both R and D are roughly constant and if T is primarily generated by the two parents (i.e., secondary conjugation by T is negligible), the conjugation efficiency can be calculated as: $\eta \approx \frac{T}{\mathit{DR}\Delta t}$. In past studies, the conjugation readout is often described as the relative frequency, defined as T per either parent (T/D or T/R), the relative increase, defined by T per either initial parent density (T/D_i or T/R_i), or the ratio of T between two experimental conditions (e.g. with and without antibiotics) [74]. Regardless of the terminology, comparisons between the conjugation readouts under different conditions are used to determine whether specific factors promote the observed increase in T.

As evident from (1), however, these alternative metrics can confound different factors that affect T. In particular, T is generated either through conjugation, or cell division. Thus, a change in T can also result from different experimental conditions (i.e., varying D or R initial values, Δt) or antibiotic-mediated selection (growth/death of D, R, or T). These effects should be excluded or minimized when determining if a factor (e.g. antibiotic) indeed affects the conjugation efficiency η.

Simple simulations illustrate that $\frac{T}{\mathit{DR}\Delta T}$, as the estimate of η, is the most robust for quantifying efficiency compared with the other metrics (Figure 1). We assume for this example that one parent is sensitive to an antibiotic A, and the other parent and transconjugant are resistant to the antibiotic tested (Figure 1a). Upon conjugation, the sensitive parent becomes resistant to the antibiotic. When the true conjugation efficiency (η) is independent of antibiotic concentration (Figure 1b), the alternative metrics can either over- or under-estimate η, depending on whether the donor, recipient, or initial density is used for the rate calculation. This bias is magnified with greater mating times (Δt). Instead, $\frac{T}{\mathit{DR}\Delta T}$ closely reflects the true conjugation efficiency (η) even in the presence of growth (Figure 1b, iv), though it is most accurate with small Δt. When the true conjugation efficiency increases 100-fold with antibiotic (Figure 1c) other metrics may falsely represent the effect of antibiotic depending on how the efficiency is defined and the sensitivities of each strain to the antibiotic.

Figure 1:

Challenges in quantifying conjugation efficiency.

a. Growth rates of three populations, the donors (D, blue), recipients (R, red), and transconjugants (T, green). The x-axis represents the concentration of a generic antibiotic to which R is sensitive, normalized with respect to the IC₅₀ value, , and both D and T are resistant to. The y-axis is growth rate of the three populations. This arbitrary selection regime was chosen as a representative sample of the potential limitations in quantification.

b. The true conjugation efficiency η does not depend on sub-inhibitory antibiotic (dotted black line). The x-axis is time, and the y-axis is the conjugation metric for four different scenarios, corresponding to the title of each panel: (i) relative frequency as defined by $\frac{T}{D}$, (ii) relative frequency as defined by $\frac{T}{R}$, (iii) relative increase as defined by $\frac{T}{P_i}$, where P is either parent since the initial density is equal, and (iv) the conjugation efficiency $\eta =\frac{T}{\mathit{DR}\Delta T}$. Red lines indicate the metric in the absence of antibiotic, and blue lines indicate the metric in the presence of antibiotic.

c. The true conjugation efficiency η increases 100-fold with the antibiotic administered (dotted line, red no antibiotic and blue with antibiotic). The x-axis is time, and the y-axis is the conjugation metric for four different scenarios described in b. Red lines indicates the metric in the absence of antibiotic, and blue lines indicate the metric in the presence of antibiotic.

Quantifying effects of antibiotics on conjugation

When interpreting effects of antibiotics on conjugation, it is critical to distinguish a change in η and that in R, D, and T. The latter can result from a combination of a potential effect of antibiotics on η and the subsequent selection dynamics [75–78]. Antibiotics may promote η directly or indirectly (Figure 2). When transfer machinery expression itself is independent of antibiotic exposure, antibiotics have been speculated to elicit a global cellular level response that indirectly stimulates or suppresses conjugation. In direct induction, antibiotics can stimulate a cascade of molecular events that result in the expression of transfer machinery [79,80].

Figure 2:

Potential ways in which antibiotics influence conjugation dynamics.

a. In the absence of antibiotics, conjugation occurs at a certain rate, and the resistant progeny grow normally but are not selected for.

b. In the presence of antibiotics, there are two proposed scenarios as to how antibiotics modulate conjugation efficiency. In the first, antibiotics are thought to indirectly modulate the conjugation efficiency. In this case, antibiotics elicit an overall global response in the cell. For these cases, it is possible that either (i) the efficiency has increased, or (ii) the efficiency has not increased compared to (a), but after a period of growth and selection all three scenarios result in the same outcome. In the next case (iii), antibiotics directly induce the expression of conjugation machinery (illustrated here by excision of resistance gene from the chromosome into a circularized plasmid). This scenario is a common mechanism for conjugative transposons and integrative conjugated elements (ICE). In that case, the conjugation efficiency increases even before selection has occurred.

Past studies on these issues have resulted in apparently contradictory conclusions (i.e. see [81]), such as the same or similar antibiotics having opposing effects depending on the experimental context. Typically, donors and recipients are mixed in the presence of an antibiotic (or other stressors), and transconjugants are measured after a period of time Δt to determine effects of the antibiotic. By measuring T/R_i, one study showed that combinatory treatment of sub-inhibitory kanamycin (Kan) and streptomycin (Strep) concentrations promoted conjugation for three different conjugative plasmids in E. coli [82]. However, critical evaluation of the results suggest that conjugation efficiency may not have likewise increased, but rather the measure T/R_i reported on the time-dependent expansion of the transconjugant population. In particular, the antibiotics could have caused significant impact on the growth rates. These differences are magnified as cells proliferate over time. Indeed, when the effects of antibiotic-mediated selection were likely minimal (during the first four hours of conjugation), there were negligible differences in the readout with and without treatment. Other studies have similarly attributed an increase in frequency to indirect induction of conjugation by the antibiotic [82–84]. Whether the conjugation efficiency also increased is less clear.

Similarly, reported suppression of conjugation may not reflect a decrease in the conjugation efficiency either. One study reported that increasing antibiotic concentrations did not increase the total number of transconjugants (T) of an ESBL plasmid after 24 hours, and sufficiently high concentrations reduced T [85]. In these experiments, a slow transconjugant growth rate could mask an increase in conjugation efficiency, appearing as an overall decrease in HGT rate.

Given these caveats, the importance of accounting for time to reduce selection effects has been recognized, although it does not completely eliminate bias in data interpretation without also accounting for growth effects. For example, one study tested tetracycline (Tc) effect on the transfer of drug resistant determinants from a complex bacterial mixture of treated sewage effluent to an E. coli strain [86]. The results showed that the shortest time for reliable detection of the transfer effects is three hours. These experiments demonstrated modest increase in transconjugant number in the presence of Tc concentrations over 100-fold lower than the MIC for recipient E. coli cells. Nevertheless, the study did not assess the effects of Tc on growth of donors given that it is hardly possible in the complex and uncharacterized microbial community from the effluent, and therefore the source of the four-fold increase in HGT events remains unknown. Consistent with prior discussion in this work, authors show that higher Tc concentrations and longer experimental times result in antibiotic selection rather of transconjugants and sometimes decrease HGT frequencies.

Nevertheless, the importance of accounting for growth kinetics for quantifying the transfer rate has been previously recognized [73,77,87–89]. In doing so, the effect of various factors on the conjugation process can be fairly compared. For example, Johnsen et al decoupled selection from growth by inhibiting growth during conjugation in micro titer wells (using auxotrophic strains and minimal media), and then eliminated the parents to selectively grow transconjugants [76]. Using a sufficiently large number of well replicates, the number of discrete transfer events (x, denoted t in the original publication) could be statistically inferred. Specifically, the distribution of transconjugant cells in individual wells after the mating period follows the Poisson distribution, and therefore x can be quantified based on the combined probability of positive (growth) and negative (no growth) wells. As x represents transfer events only, it is not indicative of the kinetic interpretation η. However, in the absence of selection, changes in x can be conclusively attributed to the experimental factor being tested. Indeed, mercury did not affect x at low concentrations, but suppressed transfer at higher concentrations.

Another study limited bacterial growth to decouple selection dynamics from conjugation and address prior limitations. Using an engineered conjugation system based on the derepressed F plasmid, Lopatkin et al. showed that 10 different antibiotics covering major classes in terms of mode of action did not increase the conjugation efficiency (determined as $\eta \approx \frac{T}{\mathit{DR}\Delta t}$) when measured in the absence of growth [77]. At high concentrations, some antibiotics even slightly reduced the conjugation efficiency. The same conclusion held for other conjugation plasmids tested. In particular, antibiotics did not promote the conjugation efficiency for nine native conjugative systems, including those of ESBL pathogens. The study further showed that the selection dynamics alone could explain the increase or decrease in the fraction of transconjugants during prolonged experiments (up to 12 hours). As such, even if antibiotics did change the conjugation efficiency, contribution of the change was negligible in comparison with the influence of selection. Both this study and that by Johnsen et al found that other factors, such as energy availability and physiological state, more strongly affect the conjugation efficiency.

Quantifying HGT in complex settings

As environments become more complex, such as natural soils, toxic water, or biofilms, quantifying both HGT, and the effect of antibiotics on HGT, becomes increasingly challenging [90,91]. Indeed, quantification of conjugation in these environments faces several hurdles, such as the inability to carefully estimate the influence of a drug on growth rates, and the involvement of spatial distribution and motility in the gene transfer process. Proper control experiments must be established to tease apart different effects of antibiotics. Without these data, reports may be misleading, such as studies suggesting antibiotics may promote the conjugation efficiency in pure and activated sludge cultures [92], and reduce the efficiency in sewage water [93]. Taken together, such reports suggest that HGT rates may differ depending on the environment. Another study showed enhanced conjugation in biofilms of plasmids containing Kan resistance, when exposed to sub-inhibitory concentrations of Kan (as opposed to imipenem, a β-lactam) [94]. However, the authors do not distinguish an effect on the conjugation efficiency from the antibiotic-induced changes on the biofilm development dynamics. Growth rate estimates were not accounted for in the quantification of conjugation, which was further confounded by interchangeable use of different metrics of conjugation. Thus, the conclusion that bacteria can “sense the antibiotics to which they are resistant”, and in return, can facilitate its horizontal spread of resistance genes, is premature. Instead, the results in these studies may reflect effects of antimicrobials on biofilm formation or community composition rather than the influence on gene swapping. New experimental approaches and quantitative measures should be implemented to determine how antibiotics affect the conjugation dynamics in these environments by taking into account compounding factors; some studies have indeed begun to move in that direction [86].

Direct induction of conjugal transfer

Despite the confounding effects of selection dynamics, the effects of antibiotics on the conjugation efficiency can be inferred from mechanistic evidence at the molecular level. For example, it has been shown that antibiotic-induced SOS response in V. cholerae and E. coli can alleviate the repression for transcription of transfer genes, and induce transfer of SXT, an ICE naturally found in V. cholerae [95]. In particular, SOS response activation of RecA protein induces autocleavage of SetR repressor and hence upregulates expression of tra genes, leading to enhanced SXT transfer through conjugation. Although only the relative frequency is reported (T/D), antibiotic influence on conjugation frequency of strains with non-cleavable SetR mutants was negligible; this result suggests specificity in the antibiotic effect on SXT transfer via the SOS pathway. Similarly, excision and conjugative transfer of CTnDOT, a conjugative transposon carrying Tc resistance, is inducible by Tc [96–98] by 100-1,000 fold [99]; in the absence of Tc there was no observed transfer [100]. Again, although the conjugation readout was reported as T/R, the authors showed that increasing Tc concentration elevates the circularized plasmid concentration in the cell and upregulates expression of the conjugation machinery [101].

In these examples, however, the precise extent to which antibiotics modulate the conjugation efficiency remains unclear. In either study, effects of selection were not accounted for when quantifying conjugation. Utilizing only one parent for quantification could create an apparently biased increase: these experiments were performed in conditions where bacterial growth likely occurred (37°C, and anywhere from one to 16 hours on filter paper). Further, comparisons between species are invalid in the first case, due to differences in experimental conditions (e.g. incubation time), as evidenced by simulations (Figure 1, comparing the change in relative frequency as a function of time).

Experimental platforms to measure conjugation

Choosing an appropriate experimental platform may facilitate quantification by reducing bias inherent in some experimental methodology. Conjugation can be quantified using selection, by choosing markers (such antibiotic or heavy metal resistance that typically reside on plasmids) that can distinguish transconjugants from parental strains [76,102–104]; colony-forming units (CFU) are used for quantification. If the antibiotic being tested is similar to the one used for plating, its effects on the conjugation process should be readily distinguishable by ensuring complete separation between conjugation and selection, as described previously. In comparison, phenotypic readouts are advantageous because higher throughput platforms can be used, such as plate readers and flow cytometry for fluorescence [105]. With proper reporters, quantification using microscopy opens doors for single-cell and microfluidics analysis [106]. Indeed, a microfluidics platform was recently used to quantify conjugation dynamics by using fluorescently labeled donors, recipients, and transconjugants [77]. However, the detection must be sensitive enough to capture events occurring at low frequencies. Nucleic acid measurements, such as qPCR, can also be used in measuring the conjugation efficiency in certain systems [107]; this type of quantification does not require transconjugant enrichment (e.g. by selection), and could circumvent biases introduced by growth.

Regardless of the platform, experiments should be designed to ensure comparable results. For example, the effects of antibiotics among different systems introduce additional compounding factors, such as different sensitivities of donors and recipients to antibiotics. It is critical to understand what concentrations were used in each particular study with respect to minimal inhibitory concentration of a drug (MIC) or its IC50 value, and whether the antibiotic tested corresponds to the resistance transferred. If different time-scales are required for different systems, this should be adjusted for when comparing efficiencies.

Effects of antibiotics on transformation

Quantification of transformation and transduction often face similar challenges as quantification of conjugation, including a lack of generally accepted metric and the confounding effects of antibiotic-mediated selection. Often the transformation efficiency is defined as the number of transformants, which is then normalized to the DNA amount or to total viable cell counts (transformation frequency). However, a quantification framework, similar to that for conjugation, may be applicable here, as it was done for analyses of transformation in Azotobacter vinelandii [108]. Transformation can be approximated by mass-action kinetics between the recipient (R) and DNA concentration (N), generating a transformant (T):

$$\frac{\mathit{dT}}{\mathit{dt}}={\eta }_R\mathit{NR},$$ (2)

where η_R is the rate constant of the transformation reaction. The transformation efficiency can be approximated as ${\eta }_R=\frac{T}{\mathit{RN}\Delta t}$ under appropriate conditions, and should be evaluated case by case (for examples see [109–111]). For example, the majority of known naturally transformable organisms have inducible competence machinery with rare exceptions (Neisseria spp and to a certain extend Helicobacter pylori) [28,112,113]. The induction of competence is strictly regulated, often involving cell density-dependent quorum sensing regulation [114,115], and only a fraction (f_C) of the bacterial population (R) becomes proficient in DNA acquisition. Therefore, the transformation efficiency can be estimated as a η_R when competent cell density, R_C = f_CR, is taken into account. Also, the DNA used in transformation assays will not always contain the sequences necessary for transformation or transformant’s survival, such as for transformation with chromosomal DNA and quantification by selective plating. In this case, only a small portion of DNA is encoding the selective marker, dramatically reducing effective DNA concentration; quantification should be adjusted accordingly. This is further complicated by various DNA properties (source, topology (circular, linear, strandedness), size and sequence distributions) and in bacteria that have sequence-specific DNA uptake machinery (i.e. see [116]).

Despite the difficulties in direct comparison of transformation efficiency, antibiotics were shown to elicit a stress response that induces or enhances expression of competence genes in specific organisms [114,117,118]. Effects of antibiotics on transformable bacteria are often measured as changed expression level of competence proteins, assayed by reporter-fusions, estimating R_C, η_R, or their combination [118–120]. Several studies revealed that certain antibiotics, in particular those affecting DNA replication or causing DNA damage, enhance transformation by increasing the fraction of competent cells, while other antibiotics have no effect on transformation efficiency [118–120]. More detailed analyses of competence induction in S. pneumoniae revealed that stress-induced replication stalling was responsible for increased transcription of competence genes clustered in vicinity of the replication origin [118]. From the reported transformation frequencies alone, however, it is unclear whether such antibiotic-induced stress affects capacity of individual competent cells to take up DNA. DNA damaging compound mitomycin C did not appear to affect competence development in other species such as Bacillus subtilis and Haemophilus influenza [121].

Effects of antibiotics on transduction

In transduction, antibiotic presence can affect two types of bacterial populations – phage producers and phage recipients. These two populations can be separated both temporally and spatially, and therefore the two effects are usually measured individually.

Antibiotics may affect HGT dynamics by promoting prophage excision and host cell lysis [80]. This effect can be measured by quantifying the increased titers of phage in culture or increased transcription of phage genes in the host cells. For example, one study showed that β-lactam-mediated induction of S. aureus prophage results in high-frequency transfer of staphylococcal pathogenicity islands [122]. FQs and agriculture antibiotic carbadox induced prophage activity in Salmonella [123,124]. Another study demonstrated that certain antibiotics significantly increased prophage induction, transcription, and production of Shiga toxin 2 (Stx₂)-carrying phage in enterohemorrhagic E. coli O104:H4 strain while other drugs strongly inhibited its production or did not have an effect [125].

In these examples, differences in growth rates due to antibiotic exposure may confound data interpretation. In the last case, supernatant was collected from cultures grown for 15 hours and population density was not accounted for [125]. Even with sub-inhibitory concentrations, comparing the fold-change between treatment conditions may unfairly bias a faster growing culture. Conversely, the effects of antibiotics that significantly increased the transduction efficiency may have been under-estimated due to this difference as well.

Antibiotics may also influence transduction by affecting the recipient strain. This process can be approximated by mass-action kinetics of the interaction between a free phage particle (P) and a recipient cell (R), generating a transductant (T)[126,127]:

$$\frac{\mathit{dT}}{\mathit{dt}}={\eta }_T\mathit{PR},$$ (3)

where η_T describes the rate constant of the transduction reaction. Similar to the metrics above, the transduction efficiency can be quantified as ${\eta }_T=\frac{T}{\mathit{PR}\Delta t}$. In most previous studies, the transduction efficiency is defined as the number of transductants that acquired a phage-associated selective marker per volume of phage lysate [128], or by the multiplicity of infection (MOI), defined by the ratio of infectious agents to the target host cells [129]. As with the previous two mechanisms, both metrics are subject to bias, as they do not account for effects of time, nor the selection environment. For example, one study showed that different biocides can either promote or reduce the transduction efficiency by including the biocide in the recovery media, affecting the recipient state [130].

Concluding remark

Examination of reported antibiotics influences on the HGT rate reveals that antibiotics-promotion of gene transfer might not be as common as previously thought. The major limitations come from various metrics that do not account for time and thus selective pressure of the antimicrobials. The question to what extent antibiotics contribute to the spread of resistance through modulating HGT warrants future studies. We emphasize the need for carefully designed quantitative studies to draw definitive conclusions, as well as standardized definitions of HGT efficiency for each mode of transfer. To this end, understanding how different selection regimens influence multiple populations engaging in HGT will enable proper quantification, especially in more complex environments either with additional populations or spatial constraints. Understanding how antibiotics modulate HGT efficiency parameters is paramount to designing effective treatment strategies aimed at minimizing the spread of resistance. Such understanding can improve the use of existing antibiotics as we wait for development of novel antimicrobials, some of which potentially even targeting HGT [131–135].

Abbreviations:

HGT

horizontal gene transfer

MGE

mobile genetic element

MOB

mobility genes

MPF

mating pair formation

CFU

colony forming unit

PFU

plaque forming unit