Intracellular Quantification of an Antibiotic Metal Complex in Single Cells of Escherichia coli Using Cryo-X-ray Fluorescence Nanoimaging

Abstract

Bacterial resistance is a major public health challenge. In Gram-negative bacteria, the synergy between multidrug efflux pumps and outer membrane impermeability determines the intracellular concentration of antibiotics. Consequently, it also dictates antibiotic activity on their respective targets. Previous research has employed spectrofluorimetry and synchrotron radiation-based DUV microscopy as tools for monitoring the accumulation of fluoroquinolone antibiotics in bacteria at population and single-cell scales, respectively. Here, we show that cryo-XRF nanoimaging allows intracellular localization and quantification of a fluoroquinolone metal complex accumulation in Escherichia coli with different efflux pump expression levels. This method offers a promising avenue for elucidating the intracellular behavior of a range of metallodrugs in bacteria and for designing novel agents with unique mechanisms of action.

Antimicrobial resistance (AMR) is set to become the world’s leading cause of death over the coming decades. In 2019, there were an estimated 4.95 million AMR-related deaths, of which 1.3 million were directly attributable to infections caused by resistant bacteria.¹ This figure is expected to rise to 10 million deaths worldwide per year by 2050.² Despite this urgency, conventional organic medicinal chemistry has failed to fill the depleted antimicrobial pipeline: systematic analyses by Butler et al. since 2011 show that, although the number of traditional and nontraditional antibacterial agents in early stage clinical development is steadily increasing, drug approvals are still disappointingly few and far between.³⁻⁸ New approaches for developing next-generation antibiotics are urgently needed. This is especially true for WHO critical priority pathogens — i.e., Gram-negative carbapenem-resistant Acinetobacter baumannii; carbapenem-resistant, and third-generation cephalosporin-resistant Enterobacteriaceae —, for which most of the antibacterial agents are typically not active.^9,10 The main factor explaining the natural resistance of these bacteria is the protective barrier offered by their two-membrane envelope, comprising an outer membrane and a cytoplasmic membrane. In particular, the outer membrane is an asymmetrical bilayer of lipopolysaccharides and phospholipids that forms a passive diffusion barrier, slowing down the permeation of all antibiotics. This accumulation barrier is reinforced by active efflux pumps, capable of recognizing a wide range of xenobiotics and expelling them throughout the bacterial envelope.¹¹ The combined effect of these two barriers results in bacteria that are resistant to almost all antibacterial agents.^12,13 To overcome limited antibiotic accumulation induced by these mechanisms, it is important to develop platforms for measuring drug accumulation and activity in isogenic bacterial strains with different permeability properties by performing quantitative measurements of intracellular antibiotic accumulation and antibiotic susceptibility testing.^14−16,22 Notably, recent methodological advances in liquid chromatography-tandem mass spectrometry (LC-MS/MS) have already enabled quantitative measurements of the intracellular accumulation of commercially available antibiotics and unrelated chemical libraries in several bacterial species, such as Escherichia coli and Pseudomonas aeruginosa. This enabled the establishment of robust correlations between the physicochemical properties and the accumulation capacities of antibacterial compounds, in relation to the permeability of the outer membrane, the presence of porins, and efflux pumps.^{14−17,20−22}

Quinolone antibiotics appeared in the early 1960s, the first examples of which had a narrow spectrum of activity and unfavorable pharmacokinetic properties. Over time, the development of new quinolones led to improved analogs with a broader spectrum and higher efficacy.²³ In particular, second-generation quinolones result from a key modification by adding a fluorine atom in position R6 of the naphthyridine ring. This modification considerably increased the activity of quinolones since almost all quinolone antibiotics are fluoroquinolones.²⁴ In addition to the fluoride substitute, other second-generation fluoroquinolones, such as norfloxacin and ciprofloxacin, have been modified by the addition of a piperazine ring in position R7 and a cyclopropyl group in position R1. The R7 piperazine ring improved potency against Gram-negative bacteria, while the cyclopropyl group enhanced the compounds’ overall activity.^25,26 This combination made ciprofloxacin (Cip) the most active of the early second-generation compounds, and the first choice used against P. aeruginosa.²⁷ Quinolone antibiotics inhibit bacterial DNA synthesis by disrupting bacterial type II topoisomerases, inhibiting the catalytic activity of DNA gyrase and topoisomerase IV.^28,29 These two enzymes are essential to bacterial physiology, as they control the chromosome coiling required for DNA synthesis. Over time, quinolone resistance has become a serious problem among many emerging resistant pathogens. Target mutations are located at the binding sites of the drugs (QRDR), and their effects are generally exacerbated by decreased outer membrane permeability and increased drug efflux.^28,29 As such, the ability of these compounds to reach critical cytoplasmic concentrations remains a key factor for their antimicrobial activity. For several years, we have employed the intrinsic fluorescence properties of FQs to investigate the impact of (i) envelope-associated mechanisms of resistance (efflux pumps and porin loss) and (ii) the physicochemical properties of these drugs in their total accumulation. On the one hand, antibiotic accumulation in bacterial populations can be precisely determined by spectrofluorimetry following cell lysis. On the other hand, synchrotron-based deep ultraviolet (DUV) imaging enables the relative quantification of the kinetics of antibiotic accumulation in single bacterial cells. This has led to the proposal of a new approach to quantifying antibiotic accumulation relative to their antibacterial activity, based on the SICAR indexes.^15,18,19 However, the diffraction-limited resolution of about a hundred nanometers of DUV microscopy³⁰ and the lack of fluorescence calibration preclude obtaining both the intracellular distribution and the quantitative measurements of antibiotics concentration into single bacterial cells. To address these limitations, we conducted a study on the accumulation of a ternary complex of copper with ciprofloxacin^31,32 (CuCip, Figure S1) in isogenic strains of E. coli expressing different levels of AcrAB — the major efflux pump in this species. We used synchrotron-based cryo X-ray fluorescence imaging (nano-XRF) to specifically map the copper intracellular distribution within the whole hydrated bacterial cells, close to its native state, at nanometer resolution. This methodology yielded a distinctive insight into the intracellular quantification and spatial distribution of the antibiotic metal complex, circumventing the necessity for chemical fixation, labeling, or mechanical manipulation of the bacterial cells. Here, we show that efflux-deficient bacteria accumulated approximately 1.3 times more CuCip than wild-type bacteria. This result was validated by using the DUV fluorescence approaches mentioned above since the spectral properties of Cip and CuCip are similar (Figure S1).

With this, we believe that nano-XRF could be used to determine the intracellular fate of metalloantibiotics — a large and understudied group of compounds that could lead to the development of a much-needed new class of antibiotics³³ —, rationalize their potency, and improve their design.

Methods

Bacterial Strains, Growth Conditions, and Chemicals

Escherichia coli strains AG100 (wild type), and AG100A (acrB::Kan derivative) are described elsewhere.¹⁸ AG100Gyr and AG100AGyr are derivative strains that carry QRDR mutations in gyrA (encoding GyrA^S83L,D87N) from an MDR clinical isolate of E. coli.³⁴E. coli ST131 strain 02 was first transduced with a P1 lysate from an E. coli BW25113 inaA::Kan strain (KEIO collection, Dharmacon Reagents). The resulting transductants containing 15% genetic linkage between gyrA and inaA — gyrA and inaA map to 50.3 and 48.6 min in the E. coli chromosome, respectively — were used for P1 transduction in AG100 to obtain AG100Gyr inaA::Kan. E. coli K-12.^35,36 The kanamycin cassette was then cured using the FLP helper plasmid pCP20 to obtain AG100Gyr ΔinaA (herein referred to as AG100Gyr). AG100Gyr was then transduced with a P1 lysate from AG100A to obtain AG100AGyr. Genomic DNA was isolated from AG100Gyr and AG100AGyr using the Wizard genomic DNA purification kit (Promega). DNA sequencing was carried out using custom synthesized primers at Eurofins to confirm the presence of the QRDR mutations.

Unless stated otherwise, bacterial strains were grown aerobically in Luria–Bertani (LB) broth at 37 °C to midexponential phase (OD₆₀₀ ∼ 0.6). All commercially available chemicals were purchased at Sigma-Aldrich.

Drug Susceptibility Assays

Minimum inhibitory concentrations (MICs) were determined by the standard 2-fold microdilution method in 96-well microplates according to the guidelines of the CLSI (http://clsi.org). Cultures were grown in Mueller Hinton II broth and approximately 2 × 10⁵ cells were distributed into each well. After an incubation of 18 h at 37 °C, bacterial growth was evaluated using the 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-phenyl-2H-tetrazolium chloride (INT) reduction test. Experiments were carried out in triplicate (n = 3) and expressed as the resulting median values.

Cell Survival after Exposure to Cip and CuCip

E. coli cells were grown and exposed to 5 μM of ciprofloxacin in the form of free Cip or CuCip as described below. Control bacteria were left untreated. Cells were then collected by centrifugation (5,000 × g, 15 min, 21 °C) and resuspended in 1 volume of 50 mM sodium phosphate buffer supplemented with 2 mM MgCl₂ (NaPi buffer, pH 7.4). Five μL of 10-fold serial dilutions were spotted on LB agar plates. The plates were incubated at 37 °C for 24 h under aerobic conditions before the efficiency of plating was evaluated.

Cip and CuCip Accumulation

Accumulation assays were performed as previously described.¹⁸ Briefly, bacteria were grown to midexponential phase, collected by centrifugation (5,000 × g, 15 min, 21 °C), and resuspended in NaPi buffer to a final density of 6 × 10⁹ Colony Forming Units (CFU) per ml. A set volume (0.8 mL) of each bacterial suspension was incubated without or with the antibiotic (e.g., ciprofloxacin at a final concentration of 5 μM in the form of free Cip or CuCip, which did not affect cell survival, in a total volume of 1 mL) for 15 min at 37 °C. Then a set volume (0.8 mL) of each condition was mixed with NaPi buffer (1.1 mL) and bacterial pellets were immediately collected by centrifugation (10,000 × g, 5 min, 4 °C).

Spectrofluorimetry

Bacterial lysis was achieved with 500 μL of 0.1 M Glycine-HCl (pH 3.0) overnight at room temperature. Bacterial debris was removed by centrifugation (10,000 × g, 15 min, 4 °C). 400 μL of each clear lysate was mixed with 600 μL of Glycine-HCl and 100 μL of each sampled lysate were placed in black half-wells of a 96-well microplate. Fluorescence emission spectra were recorded with a Spark TECAN microplate reader with the following wavelengths: excitation at 275 nm and emission collected from 300 to 500 nm. Each sample’s antibiotic fluorescence intensity (λEm ∼ 450 nm) was corrected by the tryptophan maximum intensity fluorescence (λ_Em ∼ 340 nm).¹⁸ The average number of antibiotic molecules per bacterial cell was determined as described previously.¹⁸ Data were expressed as means and standard deviations from at least three independent assays (n ≥ 3).

Time-Lapse DUV-Imaging and Image Analysis

Experiments were conducted at room temperature at the DISCO beamline of synchrotron SOLEIL.^37,38 Bacteria were grown as described above and resuspended to an OD600 of 4.8 in NaPi buffer. Bacterial pellets were recovered by centrifugation (9,000 × g, 5 min, 4 °C) from 120 μL. Pellets were immediately resuspended in 40 μL of NaPi buffer without or with the antibiotic (5 μM final). For each condition, a droplet of the bacterial suspension (0.5 μL) was deposited between two quartz coverslips (ESCO Optics) to perform time-lapse DUV-fluorescence acquisitions over 30 min with a measurement every 2 min. Bacterial cells were first positioned in brightfield before DUV excitation on an inverted microscope (DUV-modified homemade full-field Zeiss Axio Observer Z-1 equipped with a Zeiss Ultrafluar objective with 100× magnification, 1.25 NA). The antibiotic fluorescence was recorded using the synchrotron beam with an excitation wavelength of 275 nm using a dichroic mirror at 310 nm (FF310-Di01-25 × 36 SEMROCK, USA) and through two appropriated emission bandpass filters. The first one is to detect Cip or CuCip fluorescence between 420 and 480 nm (OMEGA QMAX 420-480/0607, Figure S1B). The second one is to detect tryptophan and tyrosine fluorescence between 327 and 353 nm (SEMROCK FF01–340/26). The fluorescence signals from the tryptophan and the drug were recorded sequentially with an exposure time of 30 s for each filter. Fluorescence images were recorded by an Ultraviolet-sensitive back-illuminated electron-multiplying charge-coupled device (ANDOR iKon-M 934-BU). The complete setup (microscope, stages, filters, and camera) was controlled by Micro-Manager open-source software.³⁹ Images were analyzed with a set of Python scripts developed at the DISCO beamline (https://gitlab.synchrotron-soleil.fr/disco-beamline/bacteria-drug-uv-analysis) to extract the time-course accumulation of antibiotics in each bacteria. The antibiotic fluorescence intensities (A. U.) were calculated as described previously.¹⁸ For each condition, at least 3 regions of interest (ROIs) with a minimum of 10 bacteria per ROI were selected. Experiments were independently repeated twice.

X-ray Fluorescence Nanoimaging and Image Analysis

Bacteria were grown and prepared as described for antibiotic accumulation. For each condition (control vs CuCip treatment), 2 μL of the bacterial suspension was spotted on silicon nitride membranes (Silson Ltd., England) with a size of 1.5 mm × 1.5 mm and a thickness of 500 nm before blotting to remove excess media. Membranes were plunge-frozen in liquid ethane using a Leica EM GP2 automatic plunge freezer. Vitrificated samples were transferred to a cryo-box in a Leica EM VCM vacuum cryo manipulation system and kept in liquid nitrogen until analysis. Experiments were conducted under vacuum (∼1e^–7 mbar) in cryogenic conditions on the Nano-Imaging beamline ID16A of the ESRF.⁴⁰ The X-ray excitation energy was set to 17 keV and all relevant elements were detected using their K-level emission lines. A multilayer coated fixed curvature Kirkpatrick-Baez (KB) focusing mirror system provides the nano focus (∼30 nm) and a high flux of 4.1 × 10¹¹ ph/s from 1% of the broad bandpass.⁴¹ Nano-XRF measurements were performed with a step size of 20 nm/pixel or 40 nm/pixel and an integration time of 50 ms/pixel. The XRF signal was recorded with two large solid angle detectors, based on multielement silicon drift diodes, placed at approximately 90° on either side of the incident beam. The recorded maps were then converted to mass density (ng/mm²) in all figures using the open-source software PyMCA.^42,43 The absolute calibration to the elemental areal mass density (ng/mm²) was determined by a thin film standard (AXO Dresden GmbH). To extract the total amount of element per bacteria the map of areal mass density (ng/mm²) was processed as follows: (i) The bacteria contour was manually delineated using ImageJ software on the phosphorus element areal density map. (ii) The mean intensity of pixels inside this contour was computed from the areal mass density map for the copper element. (iii) This mean intensity was subtracted from a mean background value estimated from the mean inside an area without bacteria on the same area mass density map. (iv) The total amount of copper in the bacteria (in ng) was finally obtained by multiplying this background corrected mean intensity by the area of the bacteria converted into squared millimeters. Finally, this total amount of copper is converted to the number of molecules per bacteria using the molar mass of copper (63.5 g/mol). This processing was repeated on selected cells with an average size of ∼1.5 μm for each condition (9 ≤ n ≤ 16).

To obtain 3D localization of phosphorus and copper elements, angular projection tilt series were acquired from −78° to 78° with an angular step of 6° and a step size of 40 nm. As previously described above, the recorded maps for each tilt angle were fitted using PyMCA. Then, we use a slightly modified version of the joint reconstructive iteration algorithm⁴⁴ to align the tilt series for the copper and phosphorus elements. These aligned tilt series were then reconstructed using the ART algorithm with non-negative constrain of the ASTRA toolbox⁴⁵ implemented in the TomoPy python package⁴⁶ to obtain copper and phosphate tomograms. 3D segmentation and rendering of tomograms were performed with ChimeraX.⁴⁷ First small surface blobs were removed with a “surface dust” filter with a size of 10 applied on both copper and phosphorus tomograms, then a Gaussian filter to smooth surfaces was applied with a standard deviation of 3 and 1.5 Å respectively for the phosphorus and the copper. And finally, the isosurface of the phosphorus and the copper are rendered.

Protein Aggregation Assay

Bacteria were grown to the midexponential phase. Cell lysis was achieved in NaPi buffer with two passes through a cell disrupter (One Shot Machine, CellD SARL) at 2 kbars, chilling the cell suspension to 4 °C after each pass. The lysed cell suspension was clarified by centrifugation (10,000 × g, 15 min, 4 °C) to remove unbroken cells and large cellular debris. Then, the supernatant was ultracentrifuged (100,000 × g, 60 min, 4 °C) to separate the soluble (cytoplasm and periplasm) from the insoluble (inner and outer membranes) fractions. Bicinchoninic acid (BCA) assay (Thermo Scientific) was used to determine protein concentration in the soluble fraction. These extracts (1 mg/mL) were incubated with or without 5 μM CuCip for 15 min at 37 °C. Aggregates and soluble proteins were separated by centrifugation (22,000 × g, 30 min, 4 °C). Soluble proteins were precipitated with trichloroacetic acid (TCA, 15% final). The final pellets were resuspended in 100 μL Laemmli buffer and proteins were analyzed by SDS-PAGE.

Results and Discussion

Antibacterial Activity and Accumulation of Cip and CuCip in E. coli by DUV Fluorescence Approaches

Susceptibility assays were performed using isogenic E. coli K12 strains expressing various levels of the AcrAB efflux pump: AG100 (wild type, AcrAB⁺) and AG100A (acrB::Kan, AcrAB^–). MICs of Cip and CuCip were determined according to the CLSI guidelines. Overall, bacterial susceptibility to Cip and CuCip was identical and inversely proportional to the level of AcrAB expression, as the MICs were increased 4-fold in AG100 compared to AG100A (Table 1). To ascertain a correlation between antibacterial activity and drug uptake, we examined the accumulation of the two compounds in bacterial populations (Figure 1A) and in single bacterial cells (Figures 1B and S2) by using spectrofluorimetry and microspectrofluorimetry analyses, respectively. The SICAR_(IN) index is indicative of the accumulation of a given compound in AG100A, reflecting its outer membrane permeation in the absence of active efflux.¹⁸ The results demonstrated that Cip and CuCip exhibited comparable SICAR_(IN) levels, with an average of approximately 46,300–46,700 ciprofloxacin molecules per bacterial cell after 15 min of drug exposure (Figure 1A and Supporting Information_Source Data 1). To further assess the influence of efflux on the accumulation of the two compounds, the SICAR_(EFF) index was calculated as the ratio of the average number of molecules accumulated in AG100A to that in AG100.¹⁸ Once more, Cip and CuCip demonstrated comparable SICAR_(EFF) values within the range of 1.3 to 1.4 in both population and single-cell assays (Figures 1A,B and 2, Supporting Information_Source Data 1 and Source Data 2). Together, these results demonstrate that Cip and CuCip exhibit comparable accumulation characteristics in E. coli.

Table 1. Determination of MICs in μg/mL (μM).

	AG100	AG100A	AG100Gyr	AG100AGyr
Cip	0.016 (0.048)	0.004 (0.012)	>0.256 (0.77)	>0.256 (0.77)
CuCip	0.032 (0.045)	0.008 (0.011)	>0.256 (0.36)	>0.256 (0.36)

Figure 1.

Ciprofloxacin accumulation in E. coli using DUV fluorescence. (A) Intracellular amount of antibiotic (average number of molecules per bacteria) in AG100 (AcrAB⁺) and AG100A (AcrAB^–) after a 15 min incubation with Cip or CuCip. Quantification of antibiotics in each bacterial population was determined using spectrofluorimetry as described in “Methods”. Each point represents an individual replicate. Means and standard deviations are presented from independent experiments (n > 3). An unpaired t test with Welch correction was used to compare the means in antibiotic accumulation between each pair of conditions (**p < 0.01; *p < 0.05). (B) Accumulation of Cip and CuCip was monitored in single cells of AG100 and AG100A by using time-resolved DUV-fluorescence imaging for 30 min. Images were analyzed and corrected as described in “Methods”. Fluorescence intensities (A. U.) of antibiotic accumulation in single cells at t = 15 min are presented as violin plots. The experiment was independently repeated twice (AG100, n = 107; AG100 Cip, n = 137; AG100 CuCip, n = 222; AG100A, n = 154; AG100A Cip, n = 358; AG100A CuCip, n = 343). One-way ANOVA was used to compare the means in Cip and CuCip accumulation between AG100 and AG100A (****p < 0,0001 by Dunnett’s multiple comparison test). For clarity, only significant comparisons are shown.

Figure 2.

Nano-XRF element analysis of E. coli cells upon exposure to CuCip. (A) Localization of copper in AG100 (AcrAB⁺) and AG100A (AcrAB^–) after a 15 min incubation in the absence (upper panels) and in the presence of CuCip (lower panels). The areal density quantification bar (ng/mm²) is presented. The scale bars represent 1 μm. (B) Quantification of copper (ng/bacteria) in each condition. Each point represents a single bacterium; means and standard deviations are presented for each condition (AG100, n = 16; AG100 CuCip, n = 9; AG100A, n = 9; AG100A CuCip, n = 13). One-way ANOVA was used to compare the means in CuCip accumulation between AG100 and AG100A (**p < 0.01 by Dunnett’s multiple comparison test).

Accumulation of CuCip in E. coli by XRF Nanoimaging

DUV fluorescence approaches shed light on the accumulation of FQ antibiotics within bacterial cells. While spectrofluorimetry provides insight into the average number of antibiotic molecules per bacterium in a population, this technique necessitates a bacterial lysis step that could potentially compromise the precision and accuracy of the results, through leakage or contamination of the intracellular soluble fraction. Similarly, microspectrofluorimetry on single bacterial cells does not allow for drug quantification. Gameiro et al. previously studied the behavior of different FQ complexes with copper(II) in the presence and absence of 1,10-phenanthroline (phen) and showed that only ternary copper(II):FQ:phen complexes, including CuCip, are stable under physiological conditions.³² Thus, we used CuCip, which functions as free Cip, to quantify the amount of antibiotic in single bacterial cells using nano-XRF. This method has been used to quantitatively follow the intracellular distribution of various elements in mammalian cells, including endogenous elements and exogenous metals such as metal anticancer drugs, in a label-free manner and close to their native state.⁴⁸⁻⁵³ Briefly, E. coli AG100 and AG100A were grown to exponential phase, treated with or without 5 μM CuCip for 15 min at 37 °C, and cryo-fixed before nano-XRF analysis as described in the Methods section. Cryo-XRF experiments were carried out to reveal the intracellular distribution of phosphorus (P) and copper (Cu) with a nanometric resolution (around 20 nm) close to the native state (processed areal density maps of P and Cu data sets are accessible via Zenodo, DOI:10.5281/zenodo.13645779 and raw experimental data are can be obtained from the ESRF data portal, DOI:10.15151/ESRF-ES-613521539). Vitrified samples were scanned through the focus of the nanoprobe at an X-ray energy of 17 keV with a step size of 20 or 40 nm to obtain 2D XRF maps processed to obtain quantitative elemental information (see Methods for details). In addition to 2D images, one tomographic data set on each bacterial strain was also acquired to recover the 3D distribution of the different elements (see Methods for processing details, Figure 3 and Supplementary Movies S1, S2, and S3). With this, the intracellular fate of the antibiotic metal complex was revealed under near-native cellular conditions, bypassing the uncertainty associated with metal ion leakage from the bacterial cytoplasm or cross-contamination with bacterial membranes. Phosphorus is a prevalent element in bacteria, as each nucleotide in nucleic acids contains one phosphate group. Furthermore, its distribution is homogeneous, as the bacterial nucleoid occupies the entire cytoplasmic volume. Therefore, we used phosphorus areal mass density maps to delineate the contours of the bacteria (Figure S3A,B, Supporting Information_Source Data 3) and selected bacteria with surface areas evenly distributed between 1 × 10^–6 and 2 × 10^–6 mm² around a mean of approximately 1.5 × 10^–6 mm² (Figure S3C and Supporting Information_Source Data 3). These masks were then applied to the copper areal density maps to extract quantitative data that can be compared between CuCip-treated and control bacteria (Figure 2A and Supporting Information_Source Data 3). It is estimated that an E. coli cell contains approximately 1.7 × 10⁵ copper atoms.⁵⁴ However, the signal-to-noise ratio was insufficient to allow for the detection of copper in control bacteria (Figure 2A, upper panels, and 2B; Supporting Information_Source Data 3). In the case of the CuCip-treated bacteria, the amount of copper in nanograms per bacteria was 2.21 × 10^–8 ± 4.3 × 10^–9 in AG100 versus 2.94 × 10^–8 ± 4.3 × 10^–9 in AG100A (Figure 2B), with a SICAR_(EFF) value of 1.33 (Supporting Information_Source Data 3), in good agreement with the results obtained by DUV fluorescence approaches. The 2D maps also demonstrated a nonuniform distribution of copper within the CuCip-treated bacteria (Figure 2A).

Figure 3.

Tomogram reconstruction for phosphorus and copper elements. The white and dark scale bars represent 1 μm. (A) One section of the phosphorus tomogram cropped around one bacteria. (B) Same section of the copper tomogram for the same bacteria. (C) Volumetric rendering of the bacteria shown on panels (A,B) in which the cyan and the transparent gray isosurfaces correspond to copper and phosphorus, respectively. (D,E) Volumetric rendering for two other bacteria present in the same tomogram.

To explain these fluorescent foci, we first hypothesized that they might correspond to the subcellular localization of fluoroquinolone targets, specifically the type II topoisomerases.

The double helix structure of DNA leads to major topological problems during DNA replication and transcription. In E. coli, these problems are solved by type II topoisomerases — DNA gyrase and DNA topoisomerase (topo) IV — which are essential enzymes that modify DNA topology by introducing transient double-strand breaks. Replication introduces positive supercoils at the front and precatenated DNA at the back: gyrase acts upstream of the replication fork, while topo IV removes precatenanes downstream. The action of gyrase is also essential for transcription: gyrase removes the positive supercoils upstream of the RNA polymerase.^55,56 Although the rate of supercoil insertion by a single RNA polymerase is slow compared to replication, it is much more frequent. In a standard cell with two replisomes and up to 2,000 RNA polymerases, transcription introduces more positive supercoils than replication, distributed over the entire chromosome rather than concentrated in a single region. Immunogold electron microscopy of fixed E. coli cells suggested levels of 1000–3000 molecules per cell for both GyrA and GyrB, most of them randomly distributed throughout the cytoplasm.⁵⁷ More recently, high-speed single-molecule fluorescence imaging in live E. coli cells showed that ∼300 GyrA are stably bound to the chromosome, with ∼10 enzymes enriched near each replication fork. Trapping of reaction intermediates with Cip revealed that complexes near or far from the replisome perform catalysis. In addition, analysis of nonreplicating and nontranscribing cells suggested that the majority of gyrase is involved in maintaining steady-state levels of chromosome supercoiling.⁵⁸ To test this hypothesis, we engineered derivatives of AG100 and AG100A expressing GyrA with mutations at amino acids 67 and 83, known as QRDR mutations.⁵⁹ MICs of Cip and CuCip were 16- to 64-fold higher in the resulting strains, AG100Gyr and AG100AGyr, compared to the respective parental strains (Table 1). Only AG100AGyr was then prepared and analyzed by nano-XRF as described above. Surprisingly, we observed no difference in copper accumulation and distribution in CuCip-treated AG100AGyr compared to AG100A, suggesting that the copper foci do not result from colocalization with GyrA and that intact drug targets do not affect drug accumulation (Figure S4A,B). This latter conclusion was confirmed by accumulation assays of Cip and CuCip in AG100Gyr and AG100AGyr and drug quantification by spectrofluorimetry (Figure S4C, Table 2).

Table 2. Increase of Antibiotic Accumulation in the Absence of Active Efflux Determined by Different Methods after 15 min of Drug Exposure (SICAR Index of Efflux AG100A/AG100).

	Spectrofluorimetry	Microspectrofluorimetry	Nano-XRF
Cip	1.37	1.33	N. A.
CuCip	1.25	1.48	1.33

Copper has inherent antimicrobial properties.⁶⁰ This toxicity is partly due to the production of reactive oxygen species (ROS), but also because it induces ROS-independent protein aggregation.^61,62 The possibility that CuCip induces GyrA aggregation upon interaction is ruled out by previous nano-XRF observations in the gyrA mutated strain. To independently confirm this hypothesis, the soluble and insoluble fractions of control and CuCip-treated bacteria were separated. The protein content was analyzed by SDS-PAGE, but we did not find any protein aggregates (data not shown).

An alternative hypothesis for the formation of this fluorescent foci observed in nano-XRF is that they correspond to the formation of CuCip polymers. Some insight into this hypothesis could be drawn from the analysis of the X-ray crystallographic structure of the CuCip ternary complex. From this structure, Cip behaves as a deprotonated tridentate bridging ligand, coordinating to one Cu(II) center via the pyridone oxygen and an oxygen from the carboxylate, and to an adjacent Cu(II) center via a nitrogen from the piperazine terminal ring. The Cu(II) center is ultimately penta-coordinated, revealing a slightly distorted square pyramidal geometry with the two oxygen atoms of Cip and the two nitrogen atoms of phenanthroline in equatorial positions, while the nitrogen atom of the piperazine ring is in axial position. The coordination motif facilitates the formation of a polymer chain,³² which may result in an elevated concentration of Cu in localized areas, thereby providing a potential explanation for the fluorescent foci observed in nano-XRF images. In this context, future experiments will be conducted to compare the accumulation of other fluoroquinolones, such as norfloxacin, enrofloxacin, ofloxacin, and lomefloxacin, as well as their derived ternary copper complexes, for which previous research has revealed the presence of mononuclear complexes.³²

Conclusions

Cryopreserved sample imaging offers significant benefits, such as preventing drug leakage and eliminating the need for chemical fixation. In this study, we achieved nanometric resolution in quantifying the intracellular distribution of copper within CuCip-treated bacteria using cryo nano-XRF imaging. Our findings demonstrate that both population-level and single-cell analyses yield consistent and reliable antibiotic accumulation data, showcasing the robustness and complementarity of the approaches.

It is important to note, however, that each spectroscopy and microscopy method has its own set of advantages and disadvantages. Similarly, the application of these methods is limited by the photophysical and photochemical properties of the molecules or processes under observation. In this study, three distinct analytical techniques were employed: (i) Spectrofluorimetry was employed to quantify Cip molecules in bacterial populations exposed to the antibiotic over a specified period. (ii) DUV microscopy was used to estimate the kinetics of antibiotic accumulation in living isolated bacteria. (iii) X-ray fluorescence imaging was used to quantify and localize copper bound to Cip within cryo-fixed isolated bacteria that were also exposed to the antibiotic over a given period. The choice of DUV fluorescence methodologies is contingent upon the intrinsic characteristics of ciprofloxacin, whereas X-ray fluorescence is dependent upon those of copper.

Two-dimensional nano-XRF imaging revealed a nonuniform distribution of copper, with foci observed on a fluorescent background in CuCip-treated bacteria. Further examination of the foci in three randomly selected single bacteria (one AG100 and two AG100A that were exposed to CuCip) indicated that they constitute approximately 25% of the total quantity of copper and approximately 10% of the bacterial surface area (Figure S5). Furthermore, the average diameter of the foci was estimated to be approximately 100 nm, irrespective of the genetic background (Figure S6). These foci were not discernible by DUV microscopy for two principal reasons. First, it should be noted that DUV imaging is performed using a wide-field microscope, which has a spatial resolution that is 10 times lower than that of nano-XRF. This precludes visualization of objects smaller than 100 nm. Second, the image acquisition time is 30 s due to the extremely low incoming light flux used on the DISCO beamline — but necessary to avoid UV damage to the samples. As such, the Brownian motion to which antibiotic molecules are subjected inside bacteria results in the blurring of the localization of fluorescent structures, thereby further limiting resolution. One potential solution to reduce molecular motion is the chemical or cryogenic fixation of cells. In this context, a recent study has demonstrated how ultrafast cryo-arrest can facilitate the study of a molecular signaling system whose spatiotemporal organization is highly dynamic.⁶³

Nano-XRF offers a distinctive opportunity to gain insight into the spatial distribution of elements within cells. A primary limitation of 2D nano-XRF is that the acquisition is not confocal, whereby all the volume traversed by the incoming X-ray beam can emit fluorescence concurrently at that specific (x,y) location. Consequently, the observed foci can be located anywhere in the direction perpendicular (z) to the pixel locations. To date, only XRF tomography can access the precise spatial distribution of elements. However, the acquisition of XRF images for 3D reconstruction is a time-consuming process that is currently only accessible at one beamline. Consequently, further development is required to enable correlative imaging, which would facilitate rapid and straightforward access to the spatial distribution of metalloantibiotics in conjunction with the bacterial ultrastructure. One potential avenue for achieving this is through the coupling of nano-XRF with cryo-soft X-ray tomography.⁵² Another powerful correlative approach combines the use of nanoscale secondary ion mass spectrometry (nanoSIMS) with confocal laser scanning microscopy.⁶⁴ However, one of the major challenges associated with (nano)SIMS is sample preparation: (i) SIMS is a destructive technique; (ii) it requires an ultrahigh vacuum environment, which can result in cell lysis; (iii) resin embedding is an effective method for preserving cellular structure, of fixation can lead to the leakage of the diffusible fraction of the drug.

To tackle challenges posed by the emergence and spread of multidrug-resistant bacterial infections, recent reports have highlighted the necessity to develop radically novel approaches to antibiotic discovery.^10,65,66 In the context of the failure of traditional high-throughput techniques, there has been a call to explore new chemical spaces to identify leads with novel activities that function through different mechanisms, and/or target and disrupt bacterial membranes — metalloantibiotics.^33,67−69 Notably, there is a considerable diversity of properties among metals, with an infinite number of possibilities and combinations with ligands, as the number of coordination of a metal center can range from 1 to 28.⁷⁰ In addition, metal complexes also have access to a wide range of different mechanisms of action.⁷¹ The antibacterial activity of metal complexes was first documented in the early days of modern inorganic medicinal chemistry, with the example of simple ruthenium complexes, as reported by Dwyer and colleagues in 1952.⁷² Since then, numerous detailed studies have been conducted on rhenium (Re) and ruthenium (Ru) antibacterial complexes.³³ In particular, analogs of polypyridyl Ru(II) complexes have been demonstrated to exhibit exceptionally low MICs in both Gram-positive and Gram-negative species. The intrinsic luminescent properties of these compounds, which exhibit the ability to absorb visible light and subsequently emit long-wave light in the red and near-infrared spectral regions, have been exploited to probe their intracellular accumulation and localization, as well as to identify their mechanisms of action and potential target(s) using confocal microscopy approaches.⁷³ It has been suggested that some can accumulate at ribosomes when they form polysomes, while others disrupt the bacterial membrane(s).⁷⁴⁻⁷⁹ In a recent study by Blaskovich and colleagues, the CO-ADD platform was employed to screen nearly 1,000 metal-containing compounds for antibacterial activity.⁸⁰ It should be noted that not all the active compounds are expected to exhibit intrinsic DUV-fluorescence properties. A comprehensive understanding of the quantification, localization, and spatial distribution of antibacterial drugs within bacterial cells is crucial for advancing our knowledge of these drugs and for improving their therapeutic potential. Consequently, cryo nano-XRF can be readily applied to the analysis of all metalloantibiotics.

Glossary

Abbreviations

AMR

antimicrobial resistance

Cip

ciprofloxacin

CFU

colony forming units

CLSI

Clinical & Laboratory Standards Institute

CuCip

ciprofloxacin–copper(II)–1,10-phenanthroline ternary complex

DUV

deep ultraviolet

FQ(s)

fluoroquinolone(s)

INT

2-(4-iodophenyl)-3-(4-nitrophenyl)-5-phenyl-2H-tetrazolium chloride

LC-MS/MS

liquid chromatography-tandem mass spectrometry

MDR

multidrug resistant

MIC(s)

minimal inhibitory concentration(s)

ROI(s)

region(s) of interest

QRDR

quinolone resistance determining region

SICAR

structure intracellular concentration activity relationship

XRF

X-ray fluorescence

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnano.4c12664.

• SICAR index calculation from spectrofluorimetry (Source Data 1) (XLSX)

• Fluorescence monitoring in single bacteria over time by DUV imaging for all conditions (AG100 NaPi, AG100 Cip, AG100 CuCip, AG100A NaPi, AG100A Cip, and AG100A CuCip); data analysis at time 15 and 30 min; SICAR index calculation from DUV imaging (Source Data 2) (XLSX)

• Analysis of nano-XRF data for all conditions (AG100, AG100 CuCip, AG100A, AG100A CuCip, AG100AGyr, AG100AGyr CuCip); SICAR index calculation from nano-XRF (Source Data 3) (XLSX)

• The volumetric rendering of the bacteria shown in Figure 3C, in which the cyan and the transparent gray isosurfaces correspond to copper and phosphorous, respectively (Supplementary Movie 1) (MP4)

• The volumetric rendering of the bacteria shown in Figure 3D, in which the cyan and the transparent gray isosurfaces correspond to copper and phosphorous, respectively (Supplementary Movie 2) (MP4)

• Volumetric rendering of the bacteria shown in Figure 3E, in which the cyan and the transparent gray isosurfaces correspond to copper and phosphorous, respectively (Supplementary Movie 3) (MP4)

• Molecular structures and emission spectra of Cip and CuCip (Figure S1); time-resolved accumulation of Cip and CuCip by DUV-imaging and examples of antibiotic accumulation over time in individual E. coli bacteria under the DUV fluorescence microscope (Figure S2); localization and quantification of phosphorus in E. coli by nano-XRF (Figure S3); intracellular accumulation of Cip and CuCip in E. coligyrA mutants (Figure S4); analysis of the fractional mass of copper in the form of X-ray fluorescent foci in three randomly selected E. coli bacteria (Figure S5); average size of the foci (Figure S6) (PDF)

Author Contributions

^# M.D. and H.C. contributed equally. F.J. and M.M. conceived and supervised the project. M.D. performed all the experiments. F.J., M.D., and V.R. prepared the grids for ESRF. H.C. and M.D. performed experiments at SOLEIL. F.J., H.C., M.D., M.M., and V.R. performed experiments at ESRF. F.J., H.C., M.D., and M.M. performed data analysis. F.J., H.C., M.D., and M.M. wrote the manuscript. All the authors read and approved the manuscript.

M.D. was supported by a Ph.D. contract financed by SOLEIL.

The authors declare no competing financial interest.