Pathogenic Escherichia coli Possess Elevated Growth Rates under Exposure to Sub-Inhibitory Concentrations of Azithromycin

Abstract

Antimicrobial resistance (AMR) has been identified by the World Health Organization (WHO) as one of the ten major threats to global health. Advances in technology, including whole-genome sequencing, have provided new insights into the origin and mechanisms of AMR. However, our understanding of the short-term impact of antimicrobial pressure and resistance on the physiology of bacterial populations is limited. We aimed to investigate morphological and physiological responses of clinical isolates of E. coli under short-term exposure to key antimicrobials. We performed whole-genome sequencing on twenty-seven E. coli isolates isolated from children with sepsis to evaluate their AMR gene content. We assessed their antimicrobial susceptibility profile and measured their growth dynamics and morphological characteristics under exposure to varying concentrations of ciprofloxacin, ceftriaxone, tetracycline, gentamicin, and azithromycin. AMR was common, with all organisms resistant to at least one antimicrobial; a total of 81.5% were multi-drug-resistant (MDR). We observed an association between resistance profile and morphological characteristics of the E. coli over a three-hour exposure to antimicrobials. Growth dynamics experiments demonstrated that resistance to tetracycline promoted the growth of E. coli under antimicrobial-free conditions, while resistance to the other antimicrobials incurred a fitness cost. Notably, antimicrobial exposure heterogeneously suppressed bacterial growth, but sub-MIC concentrations of azithromycin increased the maximum growth rate of the clinical isolates. Our results outline complex interactions between organism and antimicrobials and raise clinical concerns regarding exposure of sub-MIC concentrations of specific antimicrobials.

1. Introduction

Antimicrobial resistance (AMR) in bacterial pathogens is a well-recognized global health problem, causing significant mortality and morbidity. In 2016, 58.6% of Escherichia coli (E. coli), 34.5% of Klebsiella pneumoniae (K. pneumoniae), 33.9% Pseudomonas aeruginosa, 55.4% of Acinetobacter spp., and 19.8% of Staphylococcus aureus isolates in Europe were resistant to at least one antimicrobial [1]. The situation in low- and middle-income countries (LMICs), such as Vietnam, may be worse, where approximately 50% of E. coli [2] and 91% of Shigella flexneri [3] isolates from clinical samples were resistant to at least three antimicrobial classes (multi-drug-resistant (MDR)). According to WHO, hospitalization costs for patients infected with drug-resistant bacteria are significantly higher than that for patients infected with susceptible bacteria [4]. Indeed, it is estimated that resistant pathogens may result in a total economic cost of USD 0.5 billion and USD 2.8 billion in Thailand and the US, respectively [5]. The outlook is bleak, as the discovery and development of new antimicrobials have declined dramatically [6]. Hence, there is an urgent need for novel approaches to comprehend how AMR organisms emerge and spread and understand how antimicrobials impact on bacterial cells.

E. coli is a member of the Enterobacteriaceae and a component of the healthy human gut microbiota. E. coli has been reported as the most prevalent pathogen associated with bacteraemia in Asia (43.8%) [2] and urinary tract infections in Africa (41.9%) [7]. In South America, diarrheagenic E. coli was responsible for 45.2% of diarrheal cases in children under five years [8]. Particularly, ETEC (enterotoxigenic E. coli) is estimated to cause approximately 222 million diarrheal cases and 51,200 deaths globally in 2016 [9]. The high prevalence and burden of E. coli has made it one of the most important pathogens and one of the best studied group of microorganisms in science. However, we have limited contemporary data on the impact of antimicrobials on MDR E. coli associated with disease. A scenario may be all too familiar in LMICs where empirical antimicrobials are administered to patients without any microbiological diagnostics or antimicrobial susceptibility testing.

Although antimicrobial susceptibility testing is the gold standard approach to identify and describe resistant bacteria, it provides no insight into the molecular and cellular response triggered by antimicrobial exposure. There have been previous investigations into the molecular mechanisms of AMR in specific bacteria [2] and their morphological [10,11,12] and physiological characteristics [13] under antimicrobial treatments. However, there are few studies directly linking molecular mechanisms of AMR with the morphology and physiology of antimicrobial-resistant organisms. Given the high burden of resistant E. coli in LMICs, we designed a holistic investigation to provide an in-depth understanding of AMR in bloodborne E. coli from both a cellular and molecular perspective using microscopic techniques and whole-genome sequencing. Additionally, we studied the growth dynamics of pathogenic E. coli to measure the physiological changes of E. coli with a diverse compendium of AMR genes upon short-term exposure to different antimicrobial classes.

2. Results

2.1. Resistome and Morphology Associated with AMR

The 27 clinical E. coli isolates exhibited a high prevalence of AMR phenotypes, with all isolates resistant to at least one antimicrobial (

Table 1. Minimum inhibition concentrations (MICs) and antimicrobial resistance of ATCC 25922 and 27 clinical isolates used in this study.

Isolate ID	Minimum Inhibition Concentrations (mg/L)
	CIP 1	AZI 1	CRO 1	TE 1	CN 1
E3585	32 (R)	48 (R)	256 (R)	256 (R)	256 (R)
E3823	32 (R)	48 (R)	256 (R)	256 (R)	256 (R)
E4569	0.064 (S)	12 (S)	32 (R)	4 (S)	1 (S)
E6348	0.38 (S)	8 (S)	256 (R)	4 (S)	256 (R)
E8964	0.38 (S)	64 (R)	256 (R)	256 (R)	64 (R)
E9833	0.38 (S)	16 (S)	256 (R)	256 (R)	1 (S)
E10085	32 (R)	48 (R)	256 (R)	16 (R)	256 (R)
E10487	32 (R)	256 (R)	256 (R)	256 (R)	96 (R)
E10996	0.094 (S)	16 (S)	256 (R)	256 (R)	1 (S)
E11030	32 (R)	4 (S)	0.094 (S)	256 (R)	64 (R)
E12236	0.25 (R)	256 (R)	1.5 (I)	256 (R)	1 (S)
E12241	0.094 (S)	8 (S)	0.064 (S)	128 (R)	1 (S)
E12674	32 (R)	6 (S)	64 (R)	128 (R)	1.5 (S)
E14252	12 (R)	16 (S)	256 (R)	256 (R)	32 (R)
E14488	32 (R)	256 (R)	256 (R)	4 (S)	16 (R)
E15475	32 (R)	256 (R)	256 (R)	256 (R)	1 (S)
E15476	32 (R)	256 (R)	256 (R)	256 (R)	1.5 (S)
E15583	32 (R)	256 (R)	256 (R)	256 (R)	1 (S)
E1456	0.25 (S)	48 (R)	256 (R)	64 (R)	256 (R)
E1965	32 (R)	256 (R)	256 (R)	4 (S)	256 (R)
E2408	32 (R)	256 (R)	256 (R)	4 (S)	64 (R)
E2542	32 (R)	256 (R)	256 (R)	4 (S)	256 (R)
E4751	32 (R)	256 (R)	256 (R)	256 (R)	256 (R)
E5306	0.38 (S)	48 (R)	256 (R)	256 (R)	256 (R)
E5610	32 (R)	256 (R)	256 (R)	4 (S)	256 (R)
E5896	0.38 (S)	256 (R)	128 (R)	256 (R)	256 (R)
E6227	32 (R)	256 (R)	256 (R)	256 (R)	1.5 (S)
ATCC25922	0.094 (S)	4 (S)	0.125 (S)	4 (S)	1.5 (S)
Susceptible	8/27 (29.6%)	8/27 (29.6%)	2/27(7.4%)	7/27 (25.9%)	10/27 (37.0%)
Intermediate	0/27 (0.0%)	0/27 (0.0%)	1/27 (3.7%)	0/27 (0.0%)	0/27 (0.0%)
Resistant	19/27 (70.4%)	19/27 (70.4%)	24/27 (88.9%)	20/27 (74.1%)	17/27 (63.0%)

¹ CIP: ciprofloxacin, AZI: azithromycin, CRO: ceftriaxone, TE: tetracycline, CN: gentamicin, S: susceptible, I: intermediate, R: resistant.

). Among those, 70.4%, 70.4%, 88.9%, 74.1%, and 64.0% were resistant to ciprofloxacin, azithromycin, ceftriaxone, tetracycline, and gentamicin, respectively (Table 1). Additionally, 22/27 (81.5%) isolates were resistant to at least three different classes of antimicrobials and defined as MDR [14].

To investigate the AMR gene profiles of the 28-isolate collection, WGS was performed on the 27 clinical E. coli isolates and a whole-genome sequence of E. coli ATCC 25922 was obtained from the NCBI database. The most prevalent sequence type (ST) was ST1193, which accounted for 8/27 isolates (29.6%); the remainder fell into 11 different STs (Figure 1b). A resistome analysis demonstrated that resistance to azithromycin and tetracycline was associated with the presence of mphA and the tet gene variants (tetA/tetR or tetB), respectively. All isolates harboring aac(3)-IIa were resistant to gentamicin. In contrast, the resistome data were insufficient for explaining ciprofloxacin resistance, which is commonly attributed to mutations in the Quinolone resistance determining region (QRDR) [15]. Eleven and eight of 27 clinical isolates, respectively, harbored bla_CTX-M-1 and bla_CTX-M-9, which, as extended-spectrum beta-lactamases, confer resistance to third-generation cephalosporins. Ceftriaxone MICs were positively correlated with the number of acquired resistance genes to beta-lactams in these isolates (i.e., the more bla resistance genes, the higher the MIC; $\rho =0.66$, p < 0.001). Moreover, three isolates (ST69 and ST410) carried bla_NDM-1, which is responsible for carbapenem resistance.

Aiming to explore the impacts of AMR and AMR genes on morphological changes, the 28 E. coli isolates (27 clinical isolates and the reference strain ATCC 25922) were subjected to antimicrobial exposure for three hours at CLSI breakpoint concentrations [16] and examined by light microscopy. The microscopy images were qualitatively classified into four principal categories: lysis, ellipsoid, rod-like shape, and elongation (Figure 1a). Figure 1b demonstrates that E. coli cells responded to antimicrobial exposure with differential modifications in their morphology, with some showing an association with a particular AMR phenotype. Specifically, ceftriaxone-susceptible and -intermediate resistant isolates were elongated, while resistant isolates retained a rod-like shape when exposed to ceftriaxone (Figure 1b). Similarly, the majority of E. coli that were resistant to azithromycin, tetracycline, and gentamicin retained a rod morphology, while the majority of susceptible organisms were ellipsoid when exposed to these antimicrobials (Figure 1b).

Upon exposed to azithromycin and tetracycline, morphological changes were more heterogeneous and not consistent with the respective susceptibility profiles of the organisms. For example, of eight tetracycline-resistant E. coli isolates, five were ellipsoid and three were rod-like shape. Ciprofloxacin exposure induced the highest level of morphological heterogeneity, in which resistant organisms were rod-like shape, elongated, or a rod–elongated mixture; ciprofloxacin-susceptible organisms were ellipsoid, rod-shaped, or lysed when exposed to ciprofloxacin. We did not observe any obvious association between AMR gene content and bacterial morphological changes under exposure to antimicrobials (Figure 1b), with the exception of the aforementioned rod morphology with mphA, aac(3)-IIa gene, and tet gene variants conferring resistance to azithromycin, gentamicin, and tetracycline, respectively.

2.2. Assessing the Growth Dynamics of E. coli under Exposure to Antimicrobials

The 28 E. coli isolates were exposed to different concentrations (0 to 256 mg/L) of ciprofloxacin, azithromycin, ceftriaxone, tetracycline, and gentamicin; growth curves were performed over 24 h. One isolate was chosen to perform three technical replicates (Figure S1) and the pair-wise correlation coefficients among three replicates were calculated. An average correlation coefficient of 0.89 was determined. Additionally, each isolate was exposed to azithromycin and tetracycline with an additional one–two biological replicates to confirm the observed effects of resistance on their growth dynamics.

More than 2500 growth curves from the 28 E. coli isolates in different antimicrobial exposure conditions were generated, and Gompertz’s, Baranyi’s, and the three-phase linear models were fit to the growth curves. Gompertz’s and Baranyi’s models fit with the lowest MSE (mean square error) and MAE (mean absolute error), and the highest $\rho$ (correlation coefficient) and $R^2$ (correlation of determination) (Figure S2). However, the performance of Gompertz’s model was superior to that of Baranyi’s model, which indicates that Gompertz’s (with three parameters) was the most suitable model for these data. Consequently, we employed Gompertz’s model to estimate the growth parameters of each curve for further analysis.

The growth features were plotted separately for five experimental conditions, and against different antimicrobial concentrations as relative to the isolate’s respective MIC (Figure 2). Generally, antimicrobials suppressed the growth of E. coli, which could be observed as the inferred $A$ reduced gradually and the area under the curve (AUC) reduced rapidly with increasing drug concentrations. Alternatively, increased antimicrobial concentrations induced a prolonged $\lambda$ in all cases. The $\mu$ against relative drug concentration was more heterogenous and largely dependent on antimicrobial class. The $\mu$ of E. coli treated with ciprofloxacin and tetracycline declined inversely with respect to an increase in the respective antimicrobial concentrations. However, when organisms were exposed to ceftriaxone, the $\mu$ remained constant, with some organisms showing increased $\mu$ at higher concentrations. Notably, under azithromycin exposure, the general trend of the 28 E. coli isolates was that $\mu$ increased in lower concentrations of the macrolide and then decreased gradually at higher drug concentrations. These data suggest that sub-inhibitory concentrations of azithromycin induce an increased rate of E. coli replication.

On further investigation, we also found that tetracycline-resistant isolates possess a fitness advantage over susceptible isolates in both antimicrobial-free conditions and at 0.5$\times$MIC (Figure 3). Specifically, at 0.5$\times$MICs of tetracycline, the $\mu$ (p < 0.0001), $A$ (p < 0.001), and AUC (p < 0.0001) of the tetracycline-resistant E. coli were all significantly higher than those of the susceptible organism (Figure 3). Without antimicrobial exposure, the AUC of tetracycline-resistant isolates was significantly higher than that of susceptible isolates (p < 0.05). Conversely, azithromycin-susceptible E. coli in drug-free conditions had a fitness advantage over azithromycin-resistant E. coli with significantly higher AUCs (p < 0.05) (Figure 3).

2.3. Effects of Azithromycin and Tetracycline Resistance on Growth Dynamics of E. coli

To further investigate the impact of azithromycin and tetracycline resistance on the growth dynamics of E. coli, we plotted the growth features of resistant and susceptible isolates at different relative concentrations of these antimicrobials. Figure 4 shows a comparison between trends of growth features (as inferred by Gompertz’s model) on data from three replicates of resistant and susceptible organisms grown in increasing concentrations of azithromycin and tetracycline. We found that the $\lambda$ of the tetracycline-resistant organism was lower than that of the susceptible organisms, while the $\mu$ and AUC of organisms resistant to tetracycline were higher than those of susceptible organisms. Conversely, the azithromycin-resistant organisms exhibited a longer $\lambda$, a higher $A$, and a lower $\mu$ than the azithromycin-susceptible organisms. Additionally, we found that the $\mu$ of both azithromycin-resistant and -susceptible E. coli peaked at low antimicrobial concentrations relative to MIC, which was supported by both Gompertz’s, Baranyi’s, and the three-phase linear models (Figure S3). These data suggest that sub-inhibitory concentrations of azithromycin actually accelerate the growth of both susceptible and resistant E. coli.

2.4. Elevation of Growth Rate in ermB^- E. coli at Sub-MIC Concentrations of Azithromycin

All azithromycin-resistant E. coli isolates (n = 19) carried mphA; additionally, seven harbored ermB and one isolate harbored lnuF. In order to investigate the influence of ermB, which was prevalent among E. coli resistant to azithromycin, on the growth features of E. coli, we compared the growth features of the E. coli with the presence (ermB⁺) and absence (ermB^-) of this gene. The $\mu$, $A$, and AUC decreased in both ermB⁺ and ermB^- isolates, while $\lambda$ was elevated with increased relative concentrations of azithromycin (Figure 5a). We found that ermB⁺ E. coli exhibited a higher $A$ and AUC, shorter $\lambda$, and lower $\mu$ than those of ermB^- E. coli. At low concentrations of azithromycin relative to MICs, $\mu$ in ermB^- isolates increased rapidly (peaked at ~0.03× MIC) and then gradually decreased. The highest $\mu$ value from the ermB⁺ organisms was lower than that from ermB^- organisms, which peaked at a greater relative concentration of azithromycin. We further found that at lower concentrations of azithromycin (absolute concentrations of 1 and 2 mg/L), the ermB^- organisms exhibited significantly higher $\mu$ than that of ermB⁺ organisms (Figure 5b), which was in concordance with data generated with relative azithromycin concentrations. Lastly, we found that MICs of ermB⁺ isolates were significantly higher than those of ermB^- isolates (p < 0.05) (Figure 5c). Our data show that MICs against azithromycin are associated with the accumulation of the macrolide resistance gene.

3. Discussion

This study systematically assessed the physiological and morphological responses of AMR E. coli to clinically relevant antimicrobials. Regarding morphology, we found a relationship between resistance to ceftriaxone and cell elongation, which has been previously reported [17,18,19,20,21,22] and also utilized to rapidly detect resistant isolates [10,11]. However, we found that the molecular basis of AMR was not associated with antimicrobial-induced morphological responses. This observation suggests that morphology-based rapid antimicrobial susceptibility testing may be performed without prior knowledge of resistance mechanisms, which is a drawback of nucleic acid-based and protein-based methods. A limitation of morphology-based testing is the highly diverse dependence of morphological changes on the types of antimicrobial, even for antimicrobials belonging to the same family. Considering the $\beta$-lactams as an example, meropenem induces a spindle morphology in E. coli, while moxalactam induces filamentous cells [23]. Thus, for each antimicrobial, a prior imaging dataset of both resistant and susceptible organisms is required to infer the susceptibility of newly isolated organisms. For other antimicrobials used in this study, the morphology–resistance correlation was less consistent, which may be addressed by additional time data provided by live-cell imaging, which may be required to build highly accurate antimicrobial susceptibility tests [11,24]. The heterogenous responses to CIP, AZI, TE, and CN might be explained by the highly diverse genetic background of the E. coli collection comprising of 11 different STs. Our results showed antimicrobial-induced morphological changes in both resistant and susceptible isolates, suggesting that bacterial cells respond to drug pressure despite the resistance to antimicrobials. This is probably due to the residue, in the cytoplasm, of active antimicrobials which is not yet extruded by efflux pumps (e.g., TetA and TetB) or inactivated by enzymes (e.g., $\beta$-lactamases). This could also be explained by the weak interaction between antimicrobials (e.g., CIP) and modified target proteins (e.g., mutated GyrA).

Due to improved manufacturing technology, the price of tetracyclines has decreased dramatically, to the extent where they are the cheapest antimicrobials available. This has made tetracycline the second most widely used antimicrobial worldwide [25]. Although the usage of tetracyclines in human health is becoming less common, they are still important antimicrobials disease prevention and growth promotion in animals [26]. Our data demonstrated a high prevalence of tetracycline-resistant E. coli via the tet gene, which is in accordance with previous studies [27]. By comparing growth features of E. coli treated with tetracycline, we found that tetracycline-resistant isolates possessed a fitness advantage over susceptible isolates. This fitness advantage in the absence of the antimicrobial may explain the ubiquity of tet gene carriage in the gut microbiota of children and the sustained persistence of tetracycline resistance [27]. Our observations concur with a pharmacodynamic model [28], in which the fitness of resistant E. coli did not differ from susceptible E. coli in antimicrobial-free conditions. However, a recent study using growth curve analysis found that tetracycline resistance may exert a fitness cost of up to 10% on E. coli compared with a wild-type strain [29].

Azithromycin has been listed by WHO as one of the critically important antimicrobial reagents for human medicine [30]. Moreover, WHO have been developing guidelines to prophylactically prescribe azithromycin to infants and children to reduce child mortality [31], which might increase the resistance against macrolide antimicrobials globally [32]. Therefore, the molecular mechanisms and prevalence of azithromycin resistance have been studied extensively [3,31,33,34,35]. However, the influence of azithromycin exposure on the physiology of bacteria is not well understood. Here, we observed an accelerated $\mu$ of both resistant and susceptible isolates under exposure to low concentrations of azithromycin, raising a concern about potential impacts of this antimicrobial on developing new resistant mutants. These results suggest further investigations are required to evaluate the potential benefits of presumptive antimicrobial treatments compared to the risk of AMR development. We speculate that at sub-MIC concentration, azithromycin molecules inhibit protein synthesis by binding to the P site on the 50S subunit of the bacterial ribosome [36], inducing up-regulation of ribosome synthesis which is associated with higher $\mu$ [37,38]. With increased concentration of the macrolide, more ribosomal molecules are bound by azithromycin, which leads to insufficient protein synthesis and reduction in $\mu$.

Notably, the trend of $\mu$ in azithromycin-resistant E. coli was lower than the trend in azithromycin-susceptible isolates. This observation may be experimental evidence supporting a mathematical model for ribosome-targeting antimicrobials, which showed that faster growth induces greater drug susceptibility [39]. Moreover, amongst the azithromycin-resistant E. coli harboring the mphA gene, the ermB⁺ isolates exhibited lower $\mu$ than that of ermB^-. We hypothesize two feasible causal pathways of how ermB influences $\mu$: (1) the presence/absence of ermB directly impacts on the $\mu$ trend or (2) the presence of ermB increases the MICs of E. coli and consequently changes the trend of $\mu$ against the relative concentration of azithromycin. The lower $\mu$ in ermB⁺ may be due to reduced ribosome synthesis as ermB allows constitutive methylations of 23S rRNA [40], which protects the complexes from azithromycin binding and potentially interferes with ribosome assembly. This suggests that the presence/absence of ermB is likely to have a direct impact on $\mu$.

This study has limitations. First, using these approaches, we were unable to develop an algorithmic tool to quantitatively measure morphological changes and to categorize cell shapes. Therefore, the cell shape classification may be subjective. Second, the relationship between morphology, growth dynamics, and cell viability was not investigated. Third, the correlation between the live cell counts, measured CFUs, and OD values, representing the density of both live and dead cells, was not verified. Although bacterial enumeration provides information on the actual growth rates of the cultured bacteria, we opted to analyze the growth of E. coli by monitoring OD as it enables high-throughput experiments with biological replications. Lastly, the 27 E. coli isolates possessed potential cross-interaction among AMR genes and high diversity in genetic background, which we were not able to address.

4. Materials and Methods

4.1. Organisms and Whole Genome Sequencing

Twenty-seven E. coli isolates were isolated from the blood samples of Vietnamese children hospitalized with sepsis in a previous study [41], which was approved by the Oxford Tropical Research Ethics Committee (OxTREC, reference number 35–16) and the Ethic Committee of Children’s Hospital 1 (reference number 73/GCN/BVND1). E. coli ATCC 25922 was included as a control in all analyses. Minimum inhibitory concentrations (MICs) against ciprofloxacin, azithromycin, ceftriaxone, tetracycline, and gentamicin were determined for all isolates by E-test (bioMerieux, Ivry-sur-Seine, France) and serial broth dilution. Whole-genome sequencing was performed on the 27 clinical isolates. DNA was extracted and purified using the Wizard Kit (Promega, Madison, WI, USA); the Nextera XT DNA Library Preparation Kit (Illumina, San Diego, CA, USA) was used to fragment DNA with adapter ligation. DNA sequencing was performed on a MiSeq platform, using MiSeq Reagent Kit V2 (Illumina, San Diego, CA, USA) to produce 250-bp paired-end reads. Sequencing data are available at www.ebi.ac.uk/ena (accession numbers from ERR4342165 to ERR4342191).

4.2. Monitoring and Modeling Bacterial Growth

For growth dynamics experiments, two-fold serial broth dilutions were conducted in 96-well U-bottom plates (Thermo Fisher Scientific, Waltham, MA, USA) to obtain differing concentrations of antimicrobials (from 0 to 512 mg/L) in 100$\mathsf{\mu }$L cation-adjusted Mueller-Hilton (MH) broth (Oxoid, Basingstoke, UK). The selected antimicrobials were ciprofloxacin, azithromycin, ceftriaxone, tetracycline, and gentamicin (Sigma Aldrich and AK Scientific, MO, USA). For each of the 28 E. coli isolates, one colony of overnight-incubated E. coli was inoculated into 5 mL Luria-Bertani (LB) broth (Basingstoke, UK) for 3 h at 37 °C with shaking to reach OD₆₀₀ ~ 0.3. The bacterial suspension was diluted 1/1000 in MH broth, and 100$\mathsf{\mu }$L of the diluted suspension was mixed into each well of the 96-well plates. The final concentration of antimicrobials in each well varied from 0 to 256 mg/mL (two-fold intervals, i.e., 0, 0.5, 1, 2, …, 256). OD₆₀₀ was measured on the plates every 15 min for 24 h at 37 °C, with shaking before each measurement, in a FLUOstar OPTIMA microplate reader (BMG Labtech, Ortenberg, Germany). The output data for each condition were curated and modified Gompertz’s (1), Baranyi’s (2), and three-phase linear (3) models [13] were fit to the data by minimizing the mean square error (MSE).

$$N\left(t\right)=N_0+\mathrm{Aexp}\left\{-\exp \left[\frac{\mu e}{A}\left(\lambda -t\right)+1\right]\right\}$$

(1)

$$N\left(t\right)=N_0+\mu C\left(t\right)-\ln \left\{1+\frac{e^{\mu C\left(t\right)}-1}{e^A}\right\}$$

(2)

$$N\left(t\right)=\{\begin{array}{ccc}\hfill N_0& , & t\le \lambda \hfill \\ \hfill N_0+\mu \left(t-\lambda \right)& , & \lambda <t<t_s\hfill \\ \hfill A& , & t\ge t_s\hfill \end{array}$$

(3)

where $N\left(t\right)$ is the cell density, measured in OD₆₀₀, at timepoint $t$, $N_0$ is the cell density at $t=0$, $A$ is the maximum cell density, $\mu$ is the maximum growth rate (per hour), $\lambda$ is the lag time (in hours), $t_s$ is the time to reach stationary phase (in hours), and $C\left(t\right)=t+\frac{1}{\mu }\ln \left[e^{-\mu t}+e^{-\mu \lambda }-e^{-\mu \left(t+\lambda \right)}\right]$. Area under the curve (AUC) was approximately estimated from the growth curves using the Reimann summation method. We termed the $A$, $\mu$, $\lambda$, and AUC as growth features of a growth curve.

4.3. Bright-Field Microscopy

In order to investigate the morphology of E. coli in response to antimicrobial exposure, isolates were inoculated on a nutrient agar (NA) plate and incubated overnight at 37 °C. For each isolate, <10 colonies were inoculated into saline buffer (0.85% NaCl) to reach OD₆₀₀ ~ 0.6. Resistant isolates were diluted 1/20 to reduce cell density in a field of view under microscopic examination. The selected antimicrobial (ciprofloxacin, azithromycin, ceftriaxone, tetracycline, and gentamicin) of its corresponding CLSI break-point concentration [16] (i.e., ceftriaxone at $4 \mathsf{\mu }$g/mL) was added to the bacterial suspension and to the thin NA plate (8 mL). Then, 100 $\mathsf{\mu }$L of bacterial suspension was inoculated onto NA plates and incubated for 3 h at 37 °C. Subsequently, small pieces of agarose pad were excised from the NA plate, placed onto microscope slides, and examined under a Nikon Ni-e upright microscope with a 100× objective lens.

4.4. Data Analysis

Data analysis was performed in R (version 3.5.3) programming language [42]. To estimate the parameters from the models, the “optim” function (the R Stats package) was used to constrainedly minimize the MSE between the actual and predicted OD values. For each isolate, the growth features were plotted against relative MICs using the “ggplot2” and “gridExtra” packages. Wilcoxon rank-sum test on boxplots was carried out on the “ggpubr” package. Multilocus sequence typing (MLST) of E. coli was inferred from the whole-genome sequencing data using the SRST2 pipeline [43] on the seven housekeeping genes, including adk, fumC, gyrB, icd, mdh, purA, and recA. Resistome analysis with AMR gene identification was performed on the SRST2 pipeline. De novo assembly of sequencing data was performed using SPAdes [44], gene annotation was completed using Prokka [45], and the pan- and core genomes were deduced using Roary [46] with a minimum 99% blastp identity. Finally, a phylogenetic tree was constructed on RAxML [47] using the alignment of the core genome from Roary’s output and was visualized using the “ggtree” package.

5. Conclusions

In conclusion, we exploited growth dynamics and bacterial morphology to study the impact of AMR in clinical E. coli isolates. The morphological responses to ceftriaxone were consistently associated with the resistance to the antimicrobial, while the responses to other antimicrobials were more heterogeneous. In the absence of the drug, tetracycline resistance confers a fitness advantage on E. coli, while the other antimicrobials exert a measurable fitness cost (lower AUC and $\mu$) on these bacteria. At low concentrations of azithromycin, $\mu$ was accelerated and azithromycin-susceptible isolates possessed higher $\mu$ than resistant isolates. We additionally observed an impact of AMR genes on the growth dynamics of bacteria, in which E. coli harboring the ermB gene exhibited lower $\mu$ than ermB^- E. coli. Our study raises a concern regarding the persistence of AMR and the development of new resistance phenotypes under antimicrobial exposure, showing the need for a better understanding of how morphological and physiological responses contribute to the AMR and tolerance in bacterial pathogens.