Tuning the Polarity of Antibiotic-Cy5 Conjugates Enables Highly Selective Labeling of Binding Sites

Graphical Abstract

A systematic study on 13 new antibiotic-Cy5 dye conjugates derived from ampicillin, erythromycin, trimethoprim, and vancomycin revealed that in particular the incorporation of Cy5 dyes of low polarity leads to significantly improved properties regarding photostability, optical brightness, and distinctiveness. In turn, the related conjugates could be identified as outstanding tools for a variety of studies including fluorescence and super-resolution spectroscopy, differentiation of Gram+/- bacteria and competition binding.

Abstract

Multidrug-resistant bacteria pose a major threat to global health, even as newly introduced antibiotics continue to lose their therapeutic value. Against this background, deeper insights into bacterial interaction with antibiotic drugs are urgently required, whereas fluorescently labeled drug conjugates can serve as highly valuable tools. Herein, the preparation and biological evaluation of 13 new fluorescent antibiotic-Cy5 dye conjugates is described, in which the tuning of the polarity of the Cy5 dye proved to be a key element to achieve highly favorable properties for various fields of application.

Introduction

To determine the precise behavior of the active agents in their related target organisms, fluorescently labeled probes have emerged as widely used tools for an immense variety of applications.¹ Among the fluorescent dyes frequently employed for labeling, tetramethylrhodamines (TAMRA), fluoresceins, cyanines, dansyl, 7-(dimethylamino)-coumarin-4-acetic acid (DMACA), 7-nitrobenzofurazan (NBD) and boron-dipyrromethene (BODIPY) derivatives are the most prominent.^{1a, 1d} The development of new and powerful probes thereby is of particularly high importance in the field of antibiotic research as the continuous development of bacterial resistance towards established drugs will pose a serious threat to global health in the coming years and decades.²

Taking a closer look at dye conjugation, the antibiotics vancomycin, ampicillin, erythromycin and trimethoprim are frequently chosen as they cover major modes of action, which include interference with bacterial cell wall, protein and DNA synthesis, respectively (Table 1).³ While conjugates of ampicillin with TAMRA⁴ and fluorescein⁵, and conjugates of erythromycin with BODIPY⁶ and fluorescein⁷ have not yet been employed for direct bacterial staining, their high affinities towards the penicillin-binding protein (PBP) and bacterial ribosomes, respectively, enabled the development of assays to measure binding kinetics. Among the four antibiotics mentioned above, vancomycin derivatives have most frequently been employed for bacterial staining. To date, conjugates with fluorescein,⁸ BODIPY,⁸ Cy5.5,⁹ Sulfo-Cy3,¹⁰ Sulfo-Cy5,¹¹ and 800-CW¹² have been reported with applications ranging from the elucidation of peptidoglycan synthesis to in vivo tracking of bacterial infections. While fluorescent conjugates of trimethoprim with NBD, DMACA and DNS¹³ have been used for bacterial staining and detection of efflux pumps, related conjugates with BODIPY,¹⁴ fluorescein,¹⁴ Atto 520¹⁵, Atto 655¹⁶, Atto 680¹⁶, Alexa 647¹⁶, Sulfo-Cy3^{16, 17} and Sulfo-Cy5¹⁷ have been applied for in vivo and in vitro protein labeling.

Table 1. Overview over selected applications, advantages (✓), and drawbacks (x) of antibiotic dye conjugates.

ampicillin	erythromycin	trimethoprim	vancomycin
previous studies: various conjugates (see text)
✓ binding kinetics^4, 5	✓ binding kinetics^6, 7	✓ bacterial staining^13 ✓ efflux pumps^13 ✓ proteinlabeling^14, 15-17	✓ selective staining^9, 11 ✓ target labeling^8 ✓ in vivo tracking^12 ✓ flow cytometry^9, 11
x bacterial staining	x bacterial staining	x antibiotic activity^13 x photostability^16	x photostability

this work: non-polar antibiotic-Cy5 conjugates
✓ bacterial staining ✓ competition binding ✓ antibiotic activity ✓ super resolution and fluorescence microscopy ✓ photostability ✓ selective staining

However, many of these known dye conjugates suffer from decisive drawbacks. While the conjugation of a large fluorescent dye most often leads to a strong loss of antibiotic activity and/or target specificity compared to the parent antibiotic, smaller dye components may impact probe activity to a lesser degree. As a consequence, the shorter wavelength region will then render such conjugates less suitable for certain applications, like in vivo imaging of cell tissue.^{1a, 1d} Another drawback of common fluorescent probes is the instability of the dye components towards irradiation – also referred to as photobleaching, which may limit the fields of application dramatically.¹⁸

Regarding previous studies in our group¹⁹ and recent literature reports²⁰ we reasoned that the polarity of the fluorescent dyes could be a factor strongly influencing the aforementioned drawbacks. In this article, and based on a comprehensive study on 13 novel antibiotic-dye conjugates, we now demonstrate that the precise tuning of such conjugates can enable far deeper insights into the interaction of the related antibiotics with their respective targets. In turn, the hitherto experienced drawbacks derived from dye conjugation are strongly mitigated.

Results and Discussion

To investigate the influence of dye polarity as a key factor on staining and optical properties, a series of antibiotic-Cy5 conjugates was prepared for the major reasons that Cy5 derivatives are known for outstanding optical sensitivity^{20a, 21} and that overall polarity can be readily adjusted by introducing or omitting polar groups such as sulfonates on the Cy5 core. An overview over the fourteen antibiotic-Cy5 conjugates included in the present study is given in Figure 1.

Figure 1

Antibiotic-Cy5 conjugates A1–A2, E1–E4, T1–T4, V1–V4 and Cy5 carboxylic acids C1–C4.

Whereas ampicillin A and vancomycin V were suitable for the direct attachment to the Cy5 dyes 1–4 through an existing amino group, erythromycin E and trimethoprim T had to be previously converted to (9S)-erythromycylamine A or modified with an aminopropyl linker (TMP-amine), respectively, to enable coupling with the dye components (see Supporting Information). Among all 16 possible combinations, the conjugates A3 and A4 were also prepared, but showed insufficient stability during lyophilization, which led us to exclude these compounds from biological evaluation. Regarding the remaining 14 antibiotic-Cy5 conjugates depicted in Figure 1, yet only one report exists on the polar vancomycin-derived Cy5 derivative V4, and another, which however does not include defined structures of the antibiotic-dye conjugates that were studied.^{11, 22}

To closely connect the sulfonate substitution pattern of our selected dyes to the polarity of the dye conjugate, LogP values were experimentally determined for A1–V4 using thin layer chromatography (see Supporting Information). In this way, it was verified that polarity steadily increases in each row of antibiotic dye conjugates, with values ranging from 5.4 to 2.1 for A1–A2, from >6.3 to 2.7 for E1–E4, from >6.3 to 2.0 for T1–T4 and from 5.3 to <1.6 for V1–V4. In addition, studies on the optical properties of the dye components C1–C4 were carried out. Besides the influence of different solvents, the presence of SDS micelles (as a surrogate for bacterial cell membranes)²³ on the fluorescence properties was investigated (see Supporting Information for absorption and fluorescence spectra). As expected, all dyes displayed a negative solvatochromic effect as well as an increase of fluorescence intensity with higher solvent polarity, whereat only DMSO was an exception.²⁴ Not surprisingly, uptake into SDS micelles, which can be monitored by a positive solvatochromic shift in SDS medium compared to an aqueous environment, decreased with higher dye polarity.^20b Importantly, the fluorescence intensities of the emission spectra (excitation at 515 nm) in aqueous compared to micellar environment were found to be very similar for the dye acids C1–C4.

On this basis, we next determined the minimal inhibitory concentrations (MIC) against four bacterial strains (Corynebacterium glutamicum ATCC 13032, C. glutamicum ATCC 13032 pERP1p45_gfp, Salmonella enterica sv. Typhimurium Ames 9274, Bacillus subtilis 168) for all 14 conjugates A1–V4, their respective dye acids C1–C4 and their parent antibiotics A, E, T and V via broth microdilution (see Supporting Information). Although this approach was demanding due to coloration of the sample solutions and bacterial staining, the results did nonetheless indicate that none of the four Cy5 acids C1–C4 exerts any effects on bacterial cell growth. In contrast, especially the unpolar dye conjugates A1–V1 exhibited antibiotic activity similar or even better (E1, T1) compared to their parent antibiotic, with T1 being the first antibiotic dye conjugate of its kind with a MIC even lower than the parent antibiotic against both strains of C. glutamicum employed.¹³ In addition, V1 with a MIC of 4 μg/mL against B. subtilis 168 displays activity similar to the most active conjugate hitherto reported.⁸ While conjugates with the slightly more polar Cy5 dye 2 still exhibited some antibiotic activity, MICs were higher especially for conjugates E2 and T2, which can be due to the intracellular target. Due to strong aggregation, conjugate V2 could not be investigated at concentrations above 40 μg/mL. Concerning the most polar conjugates based on dyes 3 and 4, remaining antibiotic activity could only be detected for conjugates V3 and V4, although effective concentrations were 10 to 50 times higher than for V1.

In the next stage, the suitability of all 14 dye conjugates for fluorescence spectroscopy was evaluated. C. glutamicum ATCC 13032 pERP1p45_gfp, was included to also closely monitor the location of bacteria by intrinsic GFP.²⁵ As a general trend and in line with the previous results from MIC determination, we observed the highest fluorescence intensities and most intensive overall staining for the conjugates A1, E1, T1 and V1, each incorporating the least polar Cy5 dye 1. Control experiments with the four Cy5 acids C1–C4 revealed only very faint staining due to unspecific binding (Figure S3).

Representative images obtained with the trimethoprim-derived conjugates T1 and T3 are depicted in Figure 2 (see Supporting Information for respective images of all further conjugates). For T1 useful staining was achieved at comparably low concentrations of 6–60 μg/mL. (Figure 2A, 2B). High concentrations of T1 (600 μg/mL), on the other hand, led to fluorescence quenching, which caused the bacteria to appear as black spots in a high-background environment (Figure 2C). Notably, similar trends regarding the applied concentrations were observed for the other conjugates.

Figure 2

Fluorescence spectroscopy using conjugates T1 and T3 in 10 mM aqueous DMSO and bacterial strain C. glutamicum ATCC 13032 pERP1p45_gfp. Variation of concentration of T1 (A–C). Staining with T1 (D) compared to GFP (E). Staining with T3 (F) compared to GFP (G).

Comparing the bacterial staining by conjugates T1 and T3, and using GFP as reference, the detrimental effect of higher dye polarity on the staining capability becomes obvious. To rule out that solvent effects are responsible for the weaker performance of the polar conjugates, solutions in water and in 10 mM DMSO in water were compared for A2, E2–E4, T2–T4 and V3–V4. These experiments did not reveal any difference with respect to fluorescence microscopy. Up to this point, especially the novel non-polar conjugates A1, E1, T1 and V1 proved to be highly successful regarding bacterial staining^{1a, 1d, 11}

For further insight, especially on target specificity and unspecific binding, we conducted competitive binding experiments with the parent antibiotics A–V using the four bacterial strains earlier employed for MIC determination and fluorescence microscopy (C. glutamicum ATCC 13032, C. glutamicum ATCC 13032 pERP1p45_gfp, S. Typhimurium Ames 9274, B. subtilis 168). In particular, the 14 conjugates A1–V4 were co-incubated with their respective parent antibiotic at varying ratios and the fluorescence intensity of the dyed samples was determined after washing with PBS. Control experiments ensured target saturation by A1–V4 at all concentrations. Similar to the observations made during fluorescence microscopy, those conjugates with the least polar dye 1 gave the best results with high fluorescence intensities and clear competitive behavior, as indicated by curve progression, for E1, T1 and V1. As already observed during microscopy experiments, fluorescence readout from ampicillin conjugates (A1, A2) was comparably low. Nonetheless, clear competitive behavior between labeled conjugates and their parent antibiotics was observed. Among the trimethoprim derivatives T1-T4, the binding curves clearly indicate decreasing bacterial uptake and/or target affinity with higher dye polarity. Among the erythromycin derivatives, competitive behavior was detected for E1–E3, whereat overall fluorescence intensity was highest for E1. Binding curves obtained from E4, on the other hand, only indicate low target binding. While V1 and V3 display competitive behavior (again with higher fluorescence intensities for V1), experiments with V2 were inconclusive. This can however be due to the aforementioned tendency of V2 to aggregate in aqueous medium, thus leading to fluorescence quenching.

Importantly, the results from the competition and the control experiments (Figure 3) further indicate only low levels of unspecific binding, especially among the least polar conjugates A1–V1, for all four bacterial strains. Although additional validation and optimization of the overall method may be required, especially the non-polar conjugates A1–V1 represent a promising and valuable tool for the future determination of binding affinities of antibiotics and antibiotic candidates to their respective targets. (see Supporting Information Figures S8 and S9).

Figure 3

(A) Fluorescence readout of B. subtilis 168 (blue), C. glutamicum ATCC 13032 (orange), C. glutamicum ATCC 13032 pERP1-p45_gfp (grey) and S. Typhimurium Ames 9274 (yellow) dyed with different ratios of V1 and vancomycin V at a combined concentration of 13.8 μmol/L. The proportion of V1 to the total concentration is given in percent. (B) Control experiments to ensure target saturation for V1 at different concentrations. Percent values correspond to the concentrations of V1 in the experiments in series (A).

Having observed the favorable properties of the novel conjugates in fluorescence microscopy, we turned to investigate their potential applicability for stimulated emission depletion microscopy (STED).²⁶ Again, and in contrast to literature,²⁷ we observed the most favorable properties for the conjugates incorporating the least polar Cy5 dye component 1, namely A1, E1, T1 and V1 (Figure 4). In particular, these conjugates displayed significantly higher emission intensities and less photobleaching, although earlier reports had suggested that higher polarity leads to less bleaching – due to decreased dye aggregation, and that dye polarity had no major impact on emission.²⁷ Further on, and in line with the competitive binding experiments (Figure 3, Figures S8 and S9), the staining patterns of the 14 conjugates in general, and of conjugates with dye 1 in particular, correlated well with the target compartments and thus further confirmed low unspecific binding. Selected images from STED microscopy are summarized in Figure 4.

Figure 4

Confocal STED microscopy images of C. glutamicum ATCC 13032 pERP1-p45_gfp stained with antibiotic-dye conjugates in 10 mM DMSO (in water). Conjugates A1–A2 (A, B), E1–E4 (C–F), T1–T4 (G–J), V1–V4 (K–N).

Among the 14 conjugates, the ampicillin-derived compounds A1 and A2 were the most difficult to detect, even though bearing the low polarity dyes 1 and 2 (Figures 4A, B). Location at the bacterial cell walls could nevertheless clearly be observed for A1 and A2 in all tested Gram-positive strains including C. glutamicum ATCC 13032 (see also Supporting Information Figure S5). For S. Typhimurium Ames 9274, which is Gram-negative, no clearly localized stain but random diffusion with no consistent spatial distribution occurred. This diffusion is however most likely due to the use of an Ames strain possessing a more permeable cell wall as well as efflux pumps.²⁸

The series of images obtained with C. glutamicum and the erythromycin-derived conjugates E1–E4 nicely show the negative effect of increasing Cy5 dye polarity on the image quality (Figures 4C–F). While being mostly located inside the bacterial cells, the growing irregular localization of the erythromycin-based conjugates (from E1 to E4) can be attributed to a relative increase of membrane interactions, compared to target binding whereat both electrostatic and hydrophobic interactions with lipid bilayers and proteins have been described for dye 4.²⁹ Besides C. glutamicum, intracellular location of E was especially prevalent in B. subtilis 168. When using Gram-negative S. Typhimurium Ames 9274, a heterogeneous staining pattern occurred, which is however again likely to result from efflux-like mechanisms caused by the use of the Ames subtype (see Supporting Information).^28a

The general distribution of trimethoprim in bacteria is commonly attributed to the fact that dihydrofolate reductase is mainly located in the bacterial membrane and to a lesser extent inside the cytoplasm.¹³ This assumption is consistent with the microscopic data from the trimethoprim-derived conjugates T1–T4 applied within this study, whereat the superior performance of the least polar conjugate T1 became again apparent (Figures 4G–J). Notably, localization at the inner cell membranes can also be increased by active efflux pumps, which constitute a bacterial defence mechanism against antibiotics.¹³

Similar to the ampicillin-derived conjugates A1 and A2 (Figures 4A, B), the vancomycin-mediated fluorescent stain by conjugates V1–V4 was located along the outer cell walls (Figures 4K–N). Further location at the poles and the central region of the bacterial cells indicated accumulation at old and new division sites, whereat the best image in the series was again obtained for the least polar conjugate V1 (Figure 4K). For Gram-negative S. Typhimurium Ames 9274, the vancomycin-based conjugates V1–V4 once more diffused into the bacterial cell, similar to the ampicillin-derived conjugates (see Supporting Information).

Besides the staining of Gram-positive C. glutamicum (Figure 3) and that of Gram-negative S. Typhimurium Ames (see Supporting Information and discussion above) the Gram-positive model organism B. subtilis was studied. This strain is characterized by a distinct rod-shaped morphology and peptidoglycan synthesis being directed along the lateral cell wall to maintain the particular shape.³⁰ More precisely, the cell wall synthesis in B. subtilis was proposed to occur in a helix-like pattern.³¹ Since vancomycin is known to interfere with an intermediate of the bacterial peptidoglycan synthesis,³² we explored whether the helical pattern could be detected with the vancomycin-based conjugates V1–V4. This has not yet been achieved with a single antibiotic dye conjugate.^{8, 31} On the basis that outstanding imaging properties were again observed for the least polar conjugate V1 (c.f. V1–V4 in Figures 5A–D), this particular conjugate further enabled the analysis of different cross sections of B. subtilis (Figure 5E). While a focus on the lateral cell wall revealed a wave-like pattern (circled in blue), longitudinal cross sections showed conjugate accumulation at regular intervals along the outer cell wall (circled in red). Staining at the cell walls and poles (old division sites) could also be observed in axial cross sections (circled in white). All in all, these cross sections nicely confirm the internal helical substructure. Moreover, this demonstrates that such precise imaging can indeed be achieved with a single conjugate and does not require a mixture of an antibiotic and its respective conjugate.⁸

Figure 5

Confocal (STED) microscopy images of B. subtilis 168 stained with antibiotic-dye conjugates V1 (A, E), V2 (B), V3 (C), V4 (D) in 10 mM DMSO in water. Focal plane centered on the lateral cell wall (blue circles), longitudinal cross sections (red circles) and axial cross sections (white circles).

A major application of antibiotic dye conjugates is the selective staining of bacteria. As experiments differentiating Gram-negative and Gram-positive bacteria are known for the polar dye conjugate V4,¹¹ which however performed considerably worse than V1 in our fluorescence microscopy and competitive binding studies, we carried out similar experiments using A1 and V1. For these experiments, Gram-negative Pseudomonas fluorescens and Gram-positive C. glutamicum ATCC 13032 were chosen. In addition to being resistant towards ampicillin A and vancomycin V, P. fluorescens also secretes the fluorescent siderophore pyoverdine.³³ As shown in Figure 6, A1 and V1 were able to stain C. glutamicum (blue) selectively, with P. fluorescens (white) appearing as dark shadow in the Cy5 channel (Figures 6B, C, E, F). In the GFP channel (Figures 6A, D), both bacteria are fluorescently labelled, as secreted pyoverdine is absorbed by C. glutamicum.

Figure 6

Mixture of P. fluorescens and C. glutamicum ATCC 13032 stained with A1 (10 μg/mL) (A, B, C) and V1 (27 μg/mL) (D,E,F). GFP channel (λ_ex,max: 470 nm; λ_em,max: 525 nm) (A, D), Cy5 channel (λ_ex,max: 640 nm; λ_em,max: 690 nm) (B, C, E, F). C. glutamicum ATCC 13032 stained both with pyoverdine secreted by P. fluorescens and conjugates (A1, V1) (blue), P. fluorescens only stained by pyoverdine (white). Expanded view of a mixture of P. fluorescens and C. glutamicum ATCC 13032, whereat P. fluorescens, not stained by A1 and V1, appears as dark spots on the background (C, F).

Due to the significant and unexpected differences in photostability, which were observed among the dye acids C1–C4 and the conjugates A1–V4, especially during STED microscopy, fluorophore bleaching was also examined more closely. For this purpose, time-resolved series of fluorescence microscopy images were recorded over a period of up to 5 min (see Supporting Information). Among the erythromycin and trimethoprim derivatives E1–E4 and T1–T4, the less polar conjugates with dyes 1 and 2 were considerably more photostable than the polar ones (Figure 7). While the conjugates of ampicillin, A1 and A2, showed strong photobleaching, the vancomycin derivatives turned out as highly stable, with V1, V2 and V3 (lifetimes>5 min) surpassing V4 (lifetime ca. 45 s). The dye acids C1–C4 generally suffered from extensive photobleaching.

Figure 7

Time points (s) at which C. glutamicum ATCC 13032 pERP1p45_gfp stained with conjugates A1–V4 or dye acids C1–C4 was no longer observable in fluorescence microcopy due to loss of intensity by photobleaching.

Regarding literature, sensitivity towards photobleaching was so far mainly attributed to the formation of aggregates, especially H-aggregates, which are characterized by plane-to-plane stacking and a blue-shifted absorption being responsible for photosensitivity.³⁴ Against this background, and the fact that polar substituents on the Cy5 dye (such as sulfonate groups) are commonly expected to decrease dye aggregation and thus photosensitivity, the above-mentioned trend may at first appear surprising. A plausible explanation however is that binding of the antibiotics to their target structures leads to an effective spatial separation of the attached Cy5 dyes so that aggregation effects of the dye become less important than its inherent “single molecule” photostability.³⁵ This assumption is further supported by the low photostability observed in dye acids C1–C4, which only bind unspecifically to the outer bacterial cell wall, as well as high relative stability of vancomycin derivatives, in which the size of the antibiotic itself can prevent dye aggregation. A second benefit resulting from selective target binding could be the integration of the dye conjugate into a more rigid environment, thereby enhancing the quantum yield as well as shielding the dye from potentially harmful influences.¹⁶

Finally, we turned to evaluate whether our novel antibiotic-dye conjugates could be further useful to distinguish bacteria from larger biological structures, at which high target selectivity and low unspecific binding of the conjugates would be a key requirement.

For this purpose, adherent HeLa cells were first infected with C. glutamicum and then stained with E1. As apparent from Figure 8A, C. glutamicum can be easily distinguished from the HeLa cells, which alone give only a weak background signal (Figure 8B). In further experiments, adherent Detroit 562 human epithelial cells were infected with bacteria, which were previously stained with conjugates A1–V1. For background visualization, the nuclei of the epithelial cells were stained with DAPI (4’,6-diamidino-2-phenylindole) and their filamentous actin with Phalloidin iFluor 555. Similar to the erythromycin conjugate E1, which is exemplarily shown in Figure 8C, all further unpolar conjugates A1, T1 and V1 showed excellent staining of the bacteria when being surrounded by larger stained tissues (see Supporting Information). The high quality of visualization that was reached by the use of conjugate E1 (Figure 8C) is thereby well comparable with the stain produced by the commonly employed GFP (Figure 8D), which however requires previous genetic modification of the bacteria. In comparison, more polar conjugates (A2, E3, T3, V3) were not as successful and were often barely detectable solely with the stain produced by the respective Cy5 component (see Supporting Information). In turn, and especially in combination with super resolution microscopy, these antibiotic-dye conjugates now offer new options for visualizing pathways inside the bacterial cell. Bacteria can be stained easily and efficiently, for both microscopy and spectroscopy in no more than a few hours.

Figure 8

HeLa cells infected with C. glutamicum ATCC 13032 and stained with 21.7 μg/mL E1 (A). HeLa cells stained with 21.7 μg/mL E1 (B). Detroit 562 cells infected with stained (22 μg/mL E1) C. glutamicum ATCC 13032 pERP1p45_gfp (C and D). Cy5 channel (λ_ex,max: 640 nm; λ_em,max: 690 nm) (A, B, C) and GFP channel (λ_ex,max: 470 nm; λ_em,max: 525 nm) (D).

Conclusions

We have shown that conjugation of the antibiotics ampicillin, erythromycin, trimethoprim and vancomycin to a Cy5 dye of low polarity leads to conjugates displaying excellent properties towards a variety of applications. Related to their outstanding performance in fluorescence and super resolution microscopy (STED), the non-polar dye conjugates showed comparably high photostability, which is particularly remarkable against the background of low dye polarity. Another major drawback attributed to low polarity dye conjugates, namely unspecific binding to cell membranes,^20b could not be observed in our experiments. Additionally, conjugates V1 and A1 were able to distinguish Gram-positive from Gram-negative bacteria, thus opening up a vast area of possible applications. Further underlining their scientific potential and expanding the cell biology toolbox, the conjugates proved to be applicable for distinguishing bacteria from larger biological structures such as epithelial cells and macrophages. Based on the overall results, the novel antibiotic dye conjugates are likely to be highly useful for a multitude of applications. These can be especially helpful regarding the development of new antibiotics necessary to combat increasingly emerging multi-drug resistant pathogens. For researchers interested in the novel antibiotic-dye conjugates and with the aim to support world-wide research in the field of antibiotics, we will provide samples within collaborative projects.

Experimental Section

Materials and methods: Solvents and reagents were obtained from commercial sources and used as received. Antibiotics were either recovered from expired drugs by extraction (ampicillin, trimethoprim) or used as received in case of expired lyophilisates (erythromycin, vancomycin-HCl). ¹H NMR and ¹³C NMR spectra were recorded on Bruker AVANCE 600 (¹H: 600 MHz; ¹³C: 151 MHz) and Bruker AVANCE 400 (¹H: 400 MHz; ¹³C: 101 MHz) spectrometers. For ¹H NMR spectra, spectra are referenced to the solvent peak, CHCl₃ (7.26 ppm), (CD₃)(CD₂H)SO (2.50 ppm), CD₂HOD (3.31 ppm). Chemical shifts are reported in parts per million (ppm). Coupling constants are given in Hertz (Hz). The following abbreviations are used for the description of signals: s (singlet), d (doublet), t (triplet), q (quartet) and m (multiplet). ¹³C NMR (DEPTQ) spectra were recorded in CDCl₃ (77.2 ppm) and in (CD₃)₂SO (39.5 ppm) using the solvent signals as standard. Chemical shifts are given in parts per million (ppm). Mass spectra were recorded using electron spray ionization (ESI) and atmospheric pressure photoionization (APPI), and a sector field mass analyzer for HRMS measurements. Analytical TLC was carried out on Merck silica gel plates using short wave (254 nm) UV light to visualize components. Silica gel (Kieselgel 60, 40–63 mm, 9666 Merck) was used for flash column chromatography. The purity of the substances was determined using an analytical HPLC (0 : 100→90 : 10 acetonitrile/water+0.1 % formic acid over 15 min, XBridge C18 3.5 μm, 2.1×50 mm). The purity of all target compounds A1–V4 was determined as >95 % directly after purification (HPLC and detection at 254 nm (A2–V4) of with the combined spectrum from 210 nm–650 nm (A1)).

The dye carboxylic acids C1–C4, as well as (9S)-erythromycylamine A and aminopropyl-functionalized trimethoprim (TMP-amine) were synthesized according to known literature procedures.^{35b, 36}

General procedure for the synthesis of Cy5 conjugates A1, T1, E1 and V1

Activation of dye acid C1 (C1-NHS): The reaction was carried out based on a previously reported procedure.^36g C1 (1 equiv) was dissolved in dichloromethane at 20 mg/mL. N,N′-disuccinimidylcarbonate (DSC) (1.1 equiv) and N,N-diisopropylethylamine (DIPEA) (10 equiv) were added and the reaction mixture was stirred at room temperature for 3 h under nitrogen atmosphere. The reaction mixture was diluted with dichloromethane, washed with water, 0.1 M HCl_(aq) and brine. Drying over Na₂SO₄ was followed by removal of the solvent under reduced pressure. The obtained activated dye acid C1-NHS was used for the conjugation without further purification.

Conjugation of the activated dye carboxylic acid C1-NHS: Reactions were carried out based on a previously reported procedure.¹² C1-NHS was redissolved in DMF (20 mg/mL). The antibiotic or antibiotic amine derivative (1.5–2.0 equiv) and base (DIPEA or pyridine, 10 equiv) were added and the reaction mixture was stirred for 12 h at 25–60 °C. The reaction course was monitored by HPLC. After completion, the solvent was removed under reduced pressure. The crude product was purified by preparative HPLC as described for each individual compound. To avoid cake formation during lyophilization, the solvent was primarily removed by rotary evaporation before lyophilization. As noticed, cake formation tends to accelerate the degradation of some antibiotic-dye conjugates. During synthesis and purification, the conjugates were kept in the dark and were afterwards stored at −32 °C. Samples for NMR spectroscopy were not reused and samples for biological testing were always prepared freshly.

Ampicillin-Cy5-conjugate (A1). The activation was conducted according to the general procedure using C1 (16.9 mg, 32.5 μmol), DSC (9.1 mg, 35.8 μmol), DIPEA (56 μL, 324 μmol) at room temperature for 3 h. The coupling reaction was conducted according to the general procedure with ampicillin (22.7 mg, 64.9 μmol) and pyridine (26 μL, 324 μmol) at 25 °C for 12 h. Purification by preparative HPLC (45 : 55→62 : 38 acetonitrile/water+0.1 % formic acid over 12 min, flow rate: 30 mL/min, ZORBAX Eclipse XDB-C8) yielded A1 (15.9 mg, 18.4 μmol, 57 %) as deep blue solid. HPLC: 210–650 nm, t_r (min)=11.06 min, purity: 99 %.¹H NMR (400 MHz, DMSO-d₆) δ 8.92 (d, J=8.2 Hz, 1H), 8.51 (d, J=8.4 Hz, 1H), 8.33 (t, J=13.0 Hz, 2H), 7.66–7.57 (m, 2H), 7.49–7.18 (m, 11H), 6.55 (t, J=12.3 Hz, 1H), 6.31 (d, J=13.9 Hz, 1H), 6.26 (d, J=13.8 Hz, 1H), 5.72 (d, J=8.3 Hz, 1H), 5.37 (dd, J=8.2, 3.9 Hz, 1H), 5.27 (d, J=3.9 Hz, 1H), 4.14–4.01 (m, 2H), 3.86 (s, 1H), 3.59 (s, 3H), 2.30–2.13 (m, 2H), 1.75–1.62 (m, 14H), 1.63–1.50 (m, 2H), 1.50 (s, 3H), 1.41–1.30 (m, 5H). ¹³C NMR (101 MHz, DMSO-d₆) δ 173.7, 173.3, 172.9, 172.2, 170.6, 170.0, 165.5, 154.5, 154.4, 143.2, 142.5, 141.54, 141.47, 138.8, 128.9, 128.8, 128.6, 127.9, 127.6, 125.8, 125.2, 125.1, 122.9, 122.8, 111.5, 103.8, 103.5, 74.1, 67.2, 64.8, 62.3, 61.9, 61.8, 57.8, 55.8, 49.32, 49.30, 43.8, 35.1, 31.8, 31.5, 27.8, 27.6 (2×CH₃), 27.5 (2×CH₃), 27.1, 26.2, 25.4. IR (KBr) (cm⁻¹): 3137 (s,br), 1647 (m), 1496 (s), 1401 (vs), 1147 (m). HRMS (ESI): calcd for C₄₈H₅₆N₅O₅S [M−FA]⁺, 814.3997; found, 814.3999.

Erythromycin-Cy5-conjugate (E1). The activation was conducted according to the general procedure using C1 (16.9 mg, 32.5 μmol), DSC (9.1 mg, 35.8 μmol) and DIPEA (56 μL, 324 μmol) at room temperature for 3 h. The coupling reaction was conducted according to the general procedure with (9S)-erythromycylamine A (35.8 mg, 48.7 μmol) and DIPEA (56 μL, 324 μmol) at 50 °C for 12 h. Purification by preparative HPLC (35 : 65→50 : 50 acetonitrile/water+0.1 % formic acid over 11 min, flow rate: 30 mL/min, ZORBAX Eclipse XDB-C8) yielded E1 as deep blue solid (16.3 mg, 13.1 μmol, 40 %). HPLC: 254 nm, t_r (min)=8.01 min, purity: 98 %.¹H NMR (600 MHz, DMSO-d₆) δ 8.41–8.27 (m, 3H), 7.62 (d, J=7.4 Hz, 2H), 7.47–7.34 (m, 4H), 7.30–7.21 (m, 2H), 6.58 (t, J=12.3 Hz, 1H), 6.28 (dd, J=22.3, 13.8 Hz, 2H), 5.03–4.94 (m, 1H), 4.82–4.75 (m, 1H), 4.14–4.03 (m, 2H), 4.02–3.95 (m, 2H), 3.73–3.50 (m, 5H), 3.47–3.38 (m, 2H), 3.23 (s, 4H), 3.09 (dd, J=14.5, 6.0 Hz, 1H), 2.88 (d, J=9.3 Hz, 1H), 2.69 (q, J=6.9 Hz, 1H), 2.59–2.51 (m, 1H), 2.27 (d, J=12.3 Hz, 9H), 2.11–2.00 (m, 5H), 1.89 (d, J=8.6 Hz, 1H), 1.83–1.74 (m, 2H), 1.74–1.29 (m, 20H), 1.29–0.82 (m, 25H), 0.81–0.73 (m, 4H). ¹³C NMR (101 MHz, DMSO-d₆) δ 173.2, 172.5, 171.3, 171.2, 164.5, 154.1, 154.0, 142.8, 142.0, 141.1, 141.0, 128.5, 128.4, 127.7, 127.2, 125.5, 124.7, 124.7, 122.5, 122.3, 121.5, 121.5, 111.0, 103.3, 103.1, 77.7, 75.8, 72.6, 70.5, 70.2, 70.1, 67.3, 64.9, 64.1, 48.0, 48.86, 48.8, 43.3, 41.3, 38.9, 36.2, 36.0, 34.9, 34.0, 31.1, 30.7, 30.4, 28.01, 27.97, 27.2 (2×CH₃), 27.0 (2×CH₃), 26.7, 26.0, 25.1, 21.5, 21.2, 21.0, 19.1, 18.6, 17.53, 17.49, 12.73, 12.68, 9.4. 4 signals missing due to overlap. IR (KBr) (cm⁻¹): 3136 (s,br), 2973 (s), 1728 (m), 1604 (s), 1400 (vs), 1148 (s). HRMS (ESI): calcd for C₆₉H₁₀₈ N₄O₁₃ [M−FA+H]²⁺, 600.3951; found, 600.3951.

Trimethoprim-Cy5-conjugate (T1). The activation was conducted according to the general procedure with C1 (42.1 mg, 81.1 μmol), DSC (22.9 mg, 89.3 μmol) and DIPEA (140 μL, 811 μmol) at room temperature for 3 h. The coupling reaction was conducted according to the general procedure with TMP-amine (72.2 mg, 162 μmol) and DIPEA (140 μL, 811 μmol) at 50 °C for 12 h. Purification by preparative HPLC (38 : 62→48 : 52 acetonitrile/water+0.1 % formic acid over 8 min, flow rate: 30 mL/min, ZORBAX Eclipse XDB-C8) yielded T1 as deep blue solid (41.2 mg, 48.8 μmol, 60 %). HPLC: 254 nm, t_r (min)=8.05 min, purity: 99 %.¹H NMR (600 MHz, DMSO-d₆) δ 8.37–8.25 (m, 2H), 8.21 (s, 1H), 7.71 (t, J=5.6 Hz, 1H), 7.61 (d, J=7.4 Hz, 2H), 7.50 (s, 1H), 7.44–7.33 (m, 4H), 7.30–7.19 (m, 2H), 6.55 (d, J=12.2 Hz, 3H), 6.30 (d, J=13.8 Hz, 1H), 6.24 (d, J=13.8 Hz, 1H), 6.10 (s, 2H), 5.72 (s, 2H), 4.08 (t, J=7.4 Hz, 2H), 3.79 (t, J=6.3 Hz, 2H), 3.69 (s, 6H), 3.57 (s, 3H), 3.51 (s, 2H), 3.20–3.13 (m, 2H), 2.04 (t, J=7.2 Hz, 2H), 1.72–1.62 (m, 16H), 1.58–1.50 (m, 2H), 1.38–1.29 (m, 2H). ¹³C NMR (101 MHz, DMSO-d₆) δ 173.2, 172.5, 171.7, 164.4, 163.1, 154.1, 154.0, 152.9 (2×C_q), 142.8, 142.0, 141.1, 141.0, 134.8, 134.5, 128.4, 128.4, 125.4, 124.74, 124.66, 122.5, 122.3, 111.08, 111.05, 106.0 (2×CH), 103.3, 103.1, 70.5, 55.9 (2×CH₃), 48.90, 48.87, 43.3, 35.8, 35.2, 32.6, 31.0, 29.9, 27.2 (2×CH₃), 27.0 (2×CH₃), 26.8, 25.7, 24.9. 3 signals missing due to overlap. IR (KBr) (cm⁻¹): 3130 (s, br), 1635 (m), 1400 (vs), 1148 (m). HRMS (ESI): calcd for C₄₈H₆₀N₇O₄ [M−FA]⁺, 798.4701; found, 798.4704.

Vanomycin-Cy5-conjguate (V1). The activation was conducted according to the general procedure with C1 (16.9 mg, 32.5 μmol), DSC (9.1 mg, 35.8 μmol) and DIPEA (56 μL, 324 μmol) at room temperature for 3 h. The coupling reaction was conducted according to the general procedure with vancomycin-HCl (72.4 mg, 48.7 μmol) and DIPEA (56 μL, 324 μmol) at 60 °C for 12 h. Purification by preparative HPLC (40 : 60→46 : 54 acetonitrile/water+0.1 % formic acid over 5 min, flow rate: 30 mL/min, ZORBAX Eclipse XDB-C8) yielded V1 as deep blue solid (41.0 mg, 20.9 μmol, 64 %). HPLC: 254 nm, t_r (min)=8.92 min, purity: 98 %.¹H NMR (600 MHz, DMSO-d₆) δ 8.61–8.54 (m, 1H), 8.39 (s, 2H), 8.35–8.27 (m, 3H), 7.87 (s, 1H), 7.60 (dd, J=11.0, 7.4 Hz, 2H), 7.56 (d, J=1.7 Hz, 1H), 7.50 (d, J=8.5 Hz, 1H), 7.48–7.21 (m, 10H), 7.18 (s, 1H), 7.05 (s, 1H), 6.85 (s, 1H), 6.79–6.66 (m, 2H), 6.62 (t, J=11.9 Hz, 2H), 6.42–6.34 (m, 3H), 6.29 (dd, J=13.8, 6.2 Hz, 2H), 6.22 (d, J=3.1 Hz, 1H), 5.74 (d, J=8.1 Hz, 1H), 5.55 (s, 1H), 5.32 (d, J=7.5 Hz, 1H), 5.22 (s, 1H), 5.15 (s, 2H), 5.12 (s, 1H), 4.95–4.82 (m, 1H), 4.72–4.61 (m, 1H), 4.52–4.42 (m, 1H), 4.38 (d, J=5.8 Hz, 1H), 4.23 (d, J=5.7 Hz, 2H), 4.18 (d, J=11.7 Hz, 1H), 4.06 (t, J=7.4 Hz, 2H), 3.67 (d, J=10.1 Hz, 1H), 3.61 (s, 3H), 3.57–3.42 (m, 2H), 3.35–3.17 (m, 3H), 3.07 (t, J=7.4 Hz, 1H), 2.31 (s, 3H), 2.19–2.12 (m, 1H), 2.08 (s, 1H), 1.99 (d, J=7.2 Hz, 2H), 1.83 (s, 3H), 1.81–1.57 (m, 15H), 1.54–1.24 (m, 9H), 1.10–0.98 (m, 3H), 0.87 (dd, J=25.9, 6.5 Hz, 6H). ¹³C NMR (101 MHz, DMSO-d₆) δ 174.9, 173.6, 172.9, 171.7, 171.3, 169.6, 169.5, 167.8, 165.7, 157.6, 156.7, 155.6, 154.5, 152.8, 152.5, 151.8, 150.5, 148.8, 143.2, 142.9, 142.1, 142.46, 141.51, 141.46, 136.2, 134.6, 132.5, 129.2, 128.8, 127.8, 127.6, 126.7, 126.0, 125.8, 125.2, 125.1, 124.7, 122.9, 122.8, 122.4, 121.9, 118.5, 116.6, 111.5, 111.4, 110.9, 107.2, 105.1, 103.8, 103.6, 102.5, 101.6, 98.0, 78.3, 77.6, 77.2, 72.0, 70.8, 63.7, 62.9, 62.3, 61.8, 58.6, 58.1, 55.4, 54.3, 54.2, 51.4, 49.3, 43.8, 41.5, 36.2, 35.8, 34.5, 34.2, 31.6, 31.2, 28.44, 28.41, 28.3, 28.2, 27.7, 27.6, 27.5 (2×CH₃), 27.4 (2×CH₃), 27.1, 25.9, 25.7, 25.4, 24.6, 23.5, 22.9, 17.9 (2×CH₃). 3 signals missing due to overlap. IR (KBr) (cm⁻¹): 3137 (s, br), 1653 (m), 1400 (vs), 1148 (m), 1108 (m). HRMS (ESI): calcd for C₉₈H₁₁₃Cl₂N₁₁O₂₅ [M−FA+H]²⁺, 956.8638; found, 956.8642.

Dye conjugates A2--V4 were synthesized according to similar procedures (see Supporting Information).

LogP determination: LogP values for conjugates A1–V4 were determined by TLC according to a previously published procedure.³⁷ Conjugates and reference substances were dissolved in CH₂Cl₂/methanol at appropriate concentrations. Commercially available RP-TLC plates (Alugram RP-18 W/UV₂₅₄, 0.15 mm silica gel C₁₈ with fluorescent indicator UV₂₅₄) were used. Plates were spotted with conjugates and reference substances and run with an eluent mixture of acetonitrile/aqueous 3-(N-morpholino)propanesulfonic acid (MOPS) buffer (100 mM, pH 7.4) (1/1 v/v). R_f values were determined manually. All measurements were done in triplicate at an ambient temperature of 25 °C. Five reference substances with previously reported logP values^{37, 38} (phenylboronic acid (1.59), 1-naphthol (2.84), benzophenone (3.18), anthracene (4.50), perylene (6.25)) were used and LogP values were plotted against reference R_f values to determine the calibration graph. R_f values of antibiotic dye conjugates were applied to the linear equation to obtain the respective LogP values.

Fluorescence properties of dye acids: Absorption spectra of dye acids C1--C4 were measured on a SPECORD 200 PLUS UV/Vis spectrometer (Analytik Jena, Jena (Germany)), Fluorescence spectra were measured on a Cary Eclipse fluorescence spectrophotometer (Agilent, Santa Clara (USA)). The dye acids C1--C4 were dissolved in the respective solvents at a concentration of 2.5 μM and measurements were carried out in quartz glass cuvettes (path length: 10 mm). Slit adjustments for fluorescence measurement were 5 nm and 5 nm (excitation and emission respectively) with the excitation wavelength set to 515 nm. For fluorescence spectra the residual absorption at 710 nm was subtracted from all absorption values to normalize the baseline.

Cell cultures: The cultures used were the Gram-positive Corynebacterium glutamicum ATCC 13032 pERP1p45_gfp, Corynebacterium glutamicum ATCC 13032, the Gram-negative Salmonella enterica serovar Typhimurium Ames 9274, the Gram-positive Bacillus subtilis 168 and the Gram-negative Pseudomonas fluorescens. Cultures for both B. subtilis and C. glutamicum ATCC 13032 were grown in brain-heart infusion medium (BHI; Oxoid Ltd.). Cultures for C. glutamicum ATCC 13032 pERP1- p45_gfp were grown in BHI with 50 μg/L kanamycin, cultures for S. Typhimurium Ames 9274 and Pseudomonas fluorescens were grown in lysogeny-broth medium (LB; 1 %/L tryptone; 0.5 %/L yeast extract; 1 %/L NaCl; agar; all from Oxoid Ltd.). Overnight cultures of C. glutamicum ATCC 13032 pERP1p45_gfp, C. glutamicum ATCC 13032, S. Typhimurium Ames 9274 and B. subtilis 168 were incubated under aerobic conditions at 37 °C. Overnight cultures of P. fluorescencs were incubated under aerobic conditions at 28 °C.

Antibiotic solutions: Antibiotics A–V, conjugates A1–V4 and dye acids C1--C4 were dissolved in 10 mM dimethyl sulfoxide (DMSO, p.a.) in sterile filtered deionized water at varying concentrations for all biological experiments.

MIC determination: MIC determination was carried out according to a literature procedure.³⁹ For the broth microdilution testing, solutions of the antibiotic, antibiotic-dye conjugate or dye acid were added to 96-well plates (Greiner Bio-One cell culture plates, 96-well). Bacterial suspensions (50 μL) (C. glutamicum ATCC 13032 pERP1p45_gfp, C. glutamicum ATCC 13032, S. Typhimurium Ames 9274, B. subtilis 168) were added to each well, resulting in 5×10⁵ colony forming units per mL. The 96-well plates were incubated overnight at 37 °C. The minimal inhibitory concentration was determined as the lowest concentration of antibiotic/antibiotic conjugate that visibly inhibits bacterial growth as observable with the unaided eye.

Preparation of bacteria for staining experiments: For staining experiments subcultures of C. glutamicum ATCC 13032 pERP1p45_gfp, C. glutamicum ATCC 13032, S. Typhimurium Ames 9274, B. subtilis 168 and P. fluorescens were inoculated from overnight cultures and incubated at 37 °C until an oD600 value of 0.4–0.6 was reached. Bacteria were washed multiple times with PBS and resuspended in PBS. The oD600 value was adjusted depending on the experiment. Bacteria were centrifuged and leftover medium was discarded.

Fluorescence and STED microscopy: The previously prepared bacteria were resuspended in antibiotic solution (500 μL) and incubated for 30 min at 37 °C. After another washing step, bacteria were suspended in PBS (10 μL) and pipetted onto microscope cover slips. Coverslips were dried for 10 min at 65 °C until all liquid had evaporated. The cover slips were then fixated onto glass slides with Mowiol mounting medium (10 μL).⁴⁰ The finished slides were left to dry fully overnight at 4 °C in the dark. For fluorescence microscopy, the inverted Axio Vert.A1 FL-LED fluorescence microscope with ZEN 2.3 lite Imaging Software was used. For STED microscopy, a STED microscopy build by Abberior instruments with the imspector software (version 14.0.30.60) was used. The STED laser was calibrated to 1.250 watts and the intensity was set to 50 % for all experiments.

Competitive binding: Solutions of antibiotic-conjugates and the respective unlabeled antibiotic at equimolar concentration were mixed at different ratios. For negative control experiments the unlabeled antibiotic portion was replaced by PBS solution. The prepared bacteria were resuspended in antibiotic sample solution (500 μL) and incubated for 30 min at 37 °C. The sample solution was discarded after centrifuging and the bacterial pellets were washed in PBS (500 μL) twice and in some cases until the supernatant appeared colorless. The washed and stained bacteria were then suspended in PBS (250 μL) and pipetted into 96-well plates (Thermo Fisher Scientific-Nunclon 96 Flat Bottom Black Polystyrene plates). Per well 200 μL of sample volume was used. A Tecan infinite 200Pro spectrophotometer with i-control 2.0.10.0 was used for fluorescence intensity measurements. The excitation wavelength was set to 640 nm and the wavelength for emission to 690 nm. The amplification was set manually to 50. Per measurement, 25 flashes were emitted.

Photostability: Series of timed images were taken during fluorescence microscopy and the histograms were analyzed. The Cy5 fluorescence was compared to the GFP fluorescence with an emphasis on the preservation of fluorescence intensity. Using an automated program process, a picture was taken every second for 20 seconds (C1--C4, A1–V4) or every 5 seconds for 5 min (T1,T2, E1, E2, V1–V4).

Staining against commercial dye background

Hela cells: 24-well plates were prepared with cover glasses, which had been sterilized in 70 % aqueous ethanol and flamed. 5.0×10⁴ HeLa cells were seeded per well in 1 mL cell medium (CapricornScientific DMEM high glucose (4.5 g/L) with L-glutamine (+10 % FBS) which does not contain any antibiotic additives. These cells were incubated overnight at 37 °C with 5 % CO₂ and 95 % humidity. The prepared bacteria were resuspended in PBS buffer (75 μL). The HeLa cells were then infected in triplicates with the prepared bacteria (20 μL) and antibiotic solutions (500 μL) were added to the wells. The infection was carried out under cell culture conditions (37 °C, 5 % CO₂, 95 % humidity, 90 min). The cells were then washed with PBS buffer (3×1 mL), fixed with 4 % pFA (500 μL) for 20 min under cell culture conditions and the cells were again washed with PBS buffer (1 mL). The cells were stained with phalloidin iFluor 555 reagent (Phalloidin iFluor 555 reagent, Abcam, Image-iT® FX signal enhancer, Life Technologies; 1 : 1000 dilution). Then, one drop of mounting medium (Invitrogen ProLong™ Gold antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI)) per cover glass was added to the microscope slides and the cover glasses were flipped onto the mounting medium. The finished slides were left to dry fully overnight at room temperature in the dark.

Detroit 562 cells: 24-well plates were prepared with cover glasses, which had been sterilized in 70 % aqueous ethanol and flamed. Detroit 562 cells (D562, human Caucasian female pharynx carcinoma adherent cells) were seeded at 5.0×10⁵ cells per well in 1 mL cell medium (CapricornScientific DMEM high glucose (4.5 g/L) with L-glutamine and sodium pyruvate (+10 % FBS) which does not contain any antibiotic additives. The cells were incubated overnight under cell culture conditions (37 °C, 5 % CO₂, 100 % humidity).

The prepared bacteria were resuspended in antibiotic solution (500 μL) and incubated for 30 min at 37 °C. The sample solution was discarded after centrifuging and the bacterial pellets were washed twice with PBS (500 μL). The washed and stained bacteria were then suspended in PBS buffer (75 μL). The D562 cells were infected in triplicates with the washed and stained bacteria (20 μL). The infection was carried out under cell culture conditions (37 °C, 5 % CO₂, 95 % humidity, 90 min). The cells were then washed and prepared for microscopy as described for Hela cells.

For a detailed description of all biological experiments, please refer to the Supporting Information.

Supporting Information

Detailed synthetic procedures and characterization for (9S)-erythromycylamine A, aminopropyl modified trimethoprim (TMP-amine) as well as dye conjugates A1–V4.^{12, 35b, 36} Characterization of dye acids C1–C4. Detailed descriptions of biological experiments mentioned in the experimental section. LogP values for dye conjugates. Absorption and fluorescence spectra for dye acids C1–C4 in various solvents. MIC values of dye conjugates A1–V4, antibiotics A–V, and dye acids C1–C4. Optimal staining concentrations for dye conjugates A1–V4 and C1–C4. Additional images from fluorescence and STED microscopy. Competitive binding experiments for A1–V4. Histograms and images from photostability experiments. Additional images from infection experiments. ¹H NMR, DEPTQ NMR and purity runs. Additional references cited within the Supporting Information.⁴¹

Acknowledgments

The authors are grateful for the support of this project by the Studienstiftung des Deutschen Volkes (F. G.), the Deutsche Forschungsgemeinschaft (DFG, GRK1910/B3) (F. G., J. K.) and the Deutsche Bundesstiftung Umwelt (DBU, AZ34713). Open Access funding enabled and organized by Projekt DEAL.

Conflict of interest

No conflict of interest to declare.

Open Research

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.