Enzyme-Activated, Chemiluminescent Siderophore-Dioxetane Probes Enable the Selective and Highly Sensitive Detection of Bacterial Pathogens.

The sensitive detection of bacterial infections is a prerequisite for their successful treatment. The use of a chemiluminescent readout was so far hampered by an insufficient probe enrichment at the pathogens. We coupled siderophore moieties, that harness the unique iron transport system of bacteria, with enzyme-activatable dioxetanes and obtained seven trifunctional probes with high signal-to-background ratios (S/B=426-859). Conjugates with efficient iron transport capability into bacteria were identified through a growth recovery assay. All ESKAPE pathogens were labelled brightly by desferrioxamine conjugates, while catechols were weaker due to self-quenching. Bacteria could also be detected inside lung epithelial cells. The best probe 8 detected 9.1×103  CFU mL-1 of S. aureus and 5.0×104  CFU mL-1 of P. aeruginosa, while the analogous fluorescent probe 10 was 205-305fold less sensitive. This qualifies siderophore dioxetane probes for the selective and sensitive detection of bacteria.

Introduction

The escalating resistance of pathogenic bacteria to clinically employed antibiotics has become a global problem in the health care sector with severe economic and medical repercussions.[ , , , , ] The timely and specific diagnosis of a bacterial infection remains a crucial factor for its tailored and successful treatment. The common procedures for detection and susceptibility testing are based on optimized microbiological methods in automated, central labs. However, a further shortening of analysis times and the establishment of simple, point‐of‐care solutions are subject to intense research efforts. A special, growing area concerns the detection of infections in vivo by molecular imaging techniques.[ , ] Radioactively‐labelled PET or SPECT tracers visualize infections at deep body sites, but the high costs and the high demands on infrastructure prevent their broad usage in everyday clinical practice.

A simple detection method, that is also amenable to point‐of‐care applications, exploits the detection of chemiluminescence in the visible light range. Such probes require no external light source and reach excellent sensitivities and signal‐to‐background (S/B) ratios.[ , ] Turn‐on chemiluminescent phenolic dioxetanes, first discovered by Schaap and co‐workers, were significantly improved recently through the introduction of an electron‐withdrawing group at the aromatic ring. This resulted in activatable probes that are brightly emissive in aqueous environments and applicable in biological systems.[ , , ] Lately, we functionalized dioxetanes for the detection of Mycobacterium tuberculosis as well as for bacterial carbapenemases and β‐lactamases.[ , , ]

However, a broad detection of both Gram‐negative and positive bacteria that does not rely on the secretion of resistance enzymes like β‐lactamases remained elusive so far. One cause for this is the hampered intracellular accumulation and activation of such probes, impeded by impervious bacterial cell walls. To alleviate this, an enhanced translocation into prokaryotes can be achieved by hijacking their siderophore‐based iron transport systems. Siderophores are low molecular mass iron chelators that are recognized by chelator‐specific transporters which translocate them over the bacterial membrane to satiate the pathogen's iron demand. Because these transport systems are exclusively found in prokaryotes and not in mammalian cells,[ , ] they have been exploited as molecular “Trojan Horses”, designed to smuggle siderophore‐conjugated antibiotics, dyes or radioactive labels inside bacterial cells.[ , ] Recently, we introduced the MECAM (1,3,5‐N,N′,N′′‐tris‐(2,3‐dihydroxybenzoyl)‐tri amino methyl benzene) and DOTAM (1,4,7,10‐tetra azacyclododecane‐1,4,7,10‐tetra acetic acid amide) cores as artificial siderophores for bacterial imaging and antibacterial therapy.[ , ] We also demonstrated the ability of gallium‐68‐labelled DOTAM derivatives to act as bacteria‐specific PET tracers in vivo. In this study, we aimed to establish bright and selective probes for the detection of a broad spectrum of clinically relevant bacterial pathogens based on a chemiluminescent mode of detection, with superior properties compared to fluorescent probes. The results of these efforts are reported below.

Results and Discussion

In order to obtain sensitive and bacteria‐specific imaging probes with broad spectrum activity, we combined three functionalities in one molecule, i.e. a siderophore vector, which is effectively internalized, an enzyme‐trigger combination with widespread occurrence in Gram‐positive and Gram‐negative bacteria that serves to release the third component, a latest generation dioxetane moiety for bright chemiluminescence (Figure 1). For the siderophore part, DOTAM‐ and MECAM‐based synthetic cores with catecholate chelators[ , ] and the natural desferrioxamine (DFO) with hydroxamate chelators were chosen and attached to the dioxetane's acrylic acid moiety.

Figure 1

Design concept for bacteria‐targeting chemiluminescent dioxetanes. A) Untargeted dioxetanes diffuse into Gram‐positive bacteria, but uptake into Gram‐negative bacteria and activation is minor or absent due to their double‐layered cell membrane. Conjugation to siderophores enables active uptake via bacterial siderophore transporters (structures PDB: 1FEP and 1FCP). Subsequent enzymatic activation via the trigger moiety (pink), followed by the self‐immolation of the excited phenolate IV yields a bright luminescence emission. B) Structural variations of the siderophores with regard to their core, the number and the chemical nature of iron‐chelating groups. OM=outer membrane, IM=inner membrane, OMR=outer membrane receptor.

For a triggered release, the two bicomponent systems β‐galactosidase & β‐galactose or quinone oxidoreductase & trimethyl lock (TML) were chosen. β‐Galactosidases occur in E. coli, A. baumannii, K. pneumoniae as well as in some Gram‐positive strains. In addition, colorimetric tests for β‐galactosidase based on the hydrolysis of 2‐nitrophenyl β‐D‐galactopyranoside (ONPG), were also found to be positive in the presence of β‐glucuronidases,[ , ] indicating a broad substrate tolerance. Secondly, the TML trigger has been successfully employed as a self‐immolative linker in antibiotic siderophore conjugates, with activities in pathogens of the so‐called ESKAPE panel (E. faecium, S. aureus, K. pneumoniae, A. baumannii, P. aeruginosa, and Enterobacter sp). The TML release is mediated by quinone oxidoreductases, which are commonly found in a range of bacteria.

We envisaged the following cascade of events (Figure 1A): After internalization of the iron‐loaded siderophore, a cleavage of the β‐glycosidic bond with subsequent elimination of the 4‐ethylphenol spacer (I) releases the phenolate dioxetane (III). In the second approach, the reduction of the trimethyl lock yields a hydroquinone that would undergo a sterically induced lactonization, eliminate the para‐aminobenzyl alcohol spacer (II) and thus release the phenolate dioxetane (III). This phenolate would swiftly decompose under elimination of 2‐adamantanone in a chemical excitation process and yield the excited benzoate ester (IV), which decays to its ground state with concurrent release of energy in the form of a green photon. Thus, bacterial uptake of the probe and its subsequent activation by the live bacteria would be detected as a chemiluminescent signal. The synthesis of the free dioxetanes 1–2, that served as controls, and the siderophore dioxetane probes 3–9 was performed as summarized in Figure 2.

Figure 2

Synthesis of dioxetanes 1–2 and siderophore dioxetane probes 3–9. A) See also Figures S1, S2 and S6. (i) K2CO3, DMF, 50 °C, 2 h, (ii) LiOH, THF:H2O (2 : 1), 50 °C, 2 h, 67 % over 2 steps, (iii) cat. methylene blue, O2, yellow light, DCM/DMF (9 : 1), 20 min, 67 % (iv) 31, iso‐butyl chloroformate, N‐methylmorpholine (NMM); THF, 0–25 °C, 2 h, then NMM, THF, 0–25 °C, 1 h, 43 %, (v) NHS, DCC, DCM, 1.5 h, (vi) 3‐azidopropan‐1‐amine, DMF, 30 min, 69 % over 2 steps, (vii) cat. methylene blue, O2, yellow light, DCM/DMF (9 : 1), 20 min, 85 %. B) See also Figures S2 and S3. (i) 21, Zn(OAc)2, DMSO/H2O (2 : 1), 23 °C, 5 min, then 20 in DMSO, 5 min, 23 °C; then sodium ascorbate/CuSO4, THPTA, 1×PBS pH 7.4, 23 °C, 1 h, 76 % (ii) 50 % TFA, anhydrous DCM, TIPS, 2 h, 25 °C, Et2O wash then centrifuge 4500 rcf, 15 min, 0 °C (iii) 23, 1 M HEPES buffer pH 7.5, DMF (1 : 1), overnight, 25 °C, 96 % over 2 steps, (iv) 2 a, MeOH, overnight, 25 °C, quantitative, (v) cat. methylene blue, DCM/DMF (1 : 9), O2, yellow light, 10 min, 85 %, (vi) 25, iso‐butyl‐chloroformate, NMM,THF, 0–23 °C, 2 h, then 24, NMM, THF, 0–23 °C, 1 h, 67 % (vii) cat. methylene blue, O2, yellow light, DCM/DMF (1 : 9), 1 % AcOH, 25 °C, 30 min, 47 %. C) See also Figures S4 and S5. (i) 26, TFA, DCM, 1 h, (ii) 16, NHS, DCC, DCM, (iii) TEA, DMF, yield over 3 steps 59 %, (iv) cat. methylene blue, O2, yellow light, DCM:DMF (9 : 1), 30 min, 84 %, (v) 28, 2‐azido acetic acid, sodium ascorbate, CuSO4, THPTA, DMSO, 1xPBS pH 7.4, 24 °C, 3 h, 1 % AcOH, 74 %, (vi) iso‐butyl chloroformate, NMM,THF, 0–23 °C, 2 h, (vii) 24, NMM, THF, 0–23 °C, 1 h, 67 % (viii) cat. methylene blue, O2, yellow light, DCM/DMF (1 : 9), 1 % AcOH, 25 °C, 30 min, 63 %. D) See also Figures S7 and S8. i) 13, DCC, NHS, 1.5 h, (ii) desferrioxamine (DFO) mesylate salt, TEA, DMF, 30 min, (iii) cat. methylene blue, O2, yellow light, DCM, 5 min, 52 % over 3 steps. (iv) 32, TEA, DMF, 40 %. For more details and substrate structures, see the Supporting Information.

In brief, the enolether 13 could be obtained by reacting phenol 11 with the benzyl bromide 12. Amide couplings at the acrylic acid moiety, combined with the oxidation of the olefin with singlet oxygen, yielded the dioxetanes 1–2 and the monocatechol dioxetane 7 over two to three synthetic steps. The DOTAM‐ and MECAM‐based dioxetanes 3–6 were synthesized either through a sequence of copper‐catalyzed azide‐alkyne cycloaddition (CuAAC) followed by mixed‐anhydride amide coupling, or a strain‐promoted azide‐alkyne cycloaddition (SPAAC), with subsequent oxidation to the dioxetane. The DFO siderophore was linked to the acrylic acid moiety of enolether 13 via its primary amine, and successive dioxygenation afforded 8. In a similar manner, the previously synthetized acrylic acid 32 was coupled to DFO to yield probe 9. In total, two free dioxetanes and seven siderophore dioxetane probes were synthetized in up to nine steps for the longest linear sequence. First, we evaluated the activation of the free and the conjugated dioxetanes 1–9. The incubation of 1–8 with NaBH4 (1 mM) or of 9 with β‐galactosidase (1.5 EU mL−1) produced a bright chemiluminescent signal, clearly distinct from the unactivated control (Figure 3, Figure S9).

Figure 3

In vitro chemiluminescence induction. A) Total light emission for quinone oxidoreductase‐triggered dioxetanes 1–2 and siderophore‐conjugates 3—8 ± 1 mM NaBH4, n=3, error bars±standard error of mean (SEM). B) Signal‐to‐background (S/B) ratios for 1–9, n=3, error bars ± SEM. C) Chemiluminescence kinetic profiles following in vitro activation of 8 in PBS at pH 7.4 ± 1 mM NaBH4. D) Chemiluminescence kinetic profiles following in vitro activation of 9 in PBS at pH 7.4 ± β‐galactosidase [1.5 EU mL−1].

The acrylic acid dioxetane 1 showed a 2–4 fold higher light emission compared to all other tested compounds (Figure 3A). However, the probes 2–9, with an acrylic amide instead of the acid, generally had a two to four ‐ fold higher signal‐to‐background (S/B) ratio than 1 (Figure 3B). As reported before, an electron‐withdrawing acrylic acid moiety at the phenoxy dioxetane drastically increased light emission compared to Schaap's probes. Consistently, a more electron‐rich acrylic amide at the same position decreased light emission while improving the probes stability. Upon addition of the activators, both DFO dioxetanes showed an immediate chemiluminescence emission, which reached its maximum after 5–10 minutes and then declined gradually (Figure 3C,D). On average, the DFO probe 8 had a similar intensity as dioxetane 2 and was 2–3 fold brighter than the catechol siderophores 3–7 (Figure 3A). According to literature data, luminol‐ as well as electro‐generated chemiluminescence can be effectively quenched by catechol or quinone addition in solution by either a concurrent radical‐mediated, a direct excited‐state, or a redox‐mediated quenching mechanism.[ , , ] We hypothesized that the reduced chemiluminescence for 3–7 might be due to intra‐ or intermolecular quenching by the siderophore's catechol groups. To test this, we incubated 1 with three different catechols (Q1 with free, Q2 with unstable, acetylated or Q3 with stable, methylated phenol moieties) in a three‐fold excess to reproduce the stoichiometry of a triscatecholate siderophore (Figure 4A). Chemiluminescence required the presence of both 1 and of the activator NaBH4 (Figure 4 and Figure S14). The highest luminescence was observed for the unquenched condition (Figure 4B, red curve). Pronounced signal reductions were observed for the addition of free catechol Q1 and, to a smaller extent, for the acetylated catechol Q2 (purple and green curves). The addition of Q3 (orange curve), resulted in a slight, non‐significantly reduced overall signal. Thus, free phenol groups, which may also be generated by acetyl cleavage from Q2, displayed more significant quenching than methyl ethers. Because an intermolecular quenching was possible, it is likely that also an intramolecular process is operative in 3–7 (potential mechanisms are outlined in Scheme S1). We also note that a similar photon‐induced electron transfer (PET) quenching effect by free catecholate was reported for the excited state of fluorescence dyes; this effect was blocked upon acylation. In summary, the data suggest a catechol‐mediated chemiluminescence quenching and thus explain the observed lower intensities for 3–7. In order to demonstrate that the TML moiety and a subsequent chemiluminescence signal in 8 can not only be triggered chemically, but also enzymatically, the probe was incubated with two bacterial NADH‐dependent quinone oxidoreductases, one from a Gram‐negative and one from a Gram‐positive species. E. coli quinone oxidoreductase 2 (QOR2) was obtained by homologous expression in E. coli, while the diaphorase from Clostridium kluyveri was commercially available (Figure S12). Methyl benzoquinone (MBQ) served as positive control substrate for both enzymes, and the obtained k cat and K m values for MBQ reduction by both enzymes were similar to previously reported values (Tables S1 and S2, Figure S11 and S13). Fitting of the data to the Michaelis Menten equation showed that diaphorase and QOR2 readily reduced probe 8 (V max of 9.5 and 19.35 μM min−1, respectively. Enzyme addition also lead to a bright luminescence signal over five hours that was significantly larger than the control in the absence of enzymes (Figures S11 B/C and S13 B/C). The higher V max, k cat and K m values for the activation of 8 with QOR2 are reflected by the increased chemiluminescence signals compared to the incubation with diaphorase. A sequence homology search (pBLAST) for QOR2 from E. coli and the commercial diaphorase yielded more than 200 sequences from various bacterial genera express similar proteins such as Klebsiella, Pseudomonas, Shigella, Salmonella, Acinetobacter, Clostridium, Staphylococcus (Supplementary Data File). This finding supports the potential for the activation of 8 by a broad range of pathogens.

Figure 4

Quenching of dioxetane chemiluminescence by catechols. A) Reagents were 1 (10 μM) in PBS at pH 7.4, ± 30 μM quencher Q1 (2,3‐dihydroxybenzoic acid), Q2 (2,3‐diacetoxybenzoic acid) or Q3 (dimethoxybenzoic acid), ± 1 mM NaBH4 as a chemical activator. B) Chemiluminescence kinetic profiles and C) summed chemiluminescence over 60 minutes in [RLU] for 10 μM probe 1 ± 30 μM quencher Q1, Q2 or Q3 and ± 1 mM NaBH4 as a chemical activator, including controls (Q +1, Q , PBS),n=3, error bars correspond to±standard error of mean (SEM). Controls are shown in Figure S9A.

Next, we investigated whether the siderophore dioxetane conjugates retained their ability to enter bacterial cells through siderophore transporters, which is a prerequisite for their activation and excitation. For this purpose, a complementation assay that measured the conjugate‐mediated delivery of ferric iron into bacteria was applied. The E. coli ΔentA and P. aeruginosa Δpvd Δpch strains cannot biosynthesize their endogenous siderophores enterobactin (ENT) or pyoverdine/pyochelin (PYO), respectively and thus are unable to grow under iron‐restricted conditions, except when a suitable (xeno‐) siderophore is added (Figure 5). The exogenous addition of the respective natural siderophores, restored the growth in both mutants. The solvent control as well as the free dioxetanes 1 and 2 did not foster bacterial growth. The DOTAM dioxetane 3 was unable to complement iron‐deficiency, and the corresponding deoxygenated enolether 3 a performed only slightly better (Figure 5A). In contrast, five conjugates (4, 6–9) restored bacterial growth efficiently in both mutant strains. Compound 5 showed a less pronounced, but still detectable growth. The robust growth recovery of 4 and 6–9 in E. coli was confirmed in P. aeruginosa (Figure 5B). The results indicate the probe's capability to shuttle ferric iron into the bacteria in the absence and also in the presence of natural siderophores, hereby proving that the necessary condition of their import is fulfilled. Given the small amounts administered as a single dose, we do not expect the probes to aggravate infections, even when applied in vivo. Probes with proven ability to transport iron into bacteria, i.e. the siderophore dioxetane conjugates 4, 6, 7, 8, 9 and control dioxetane 2, were selected to visualize ESKAPE pathogens by chemiluminescence in iron‐depleted, cation‐adjusted medium (IDCAM) in kinetic experiments over 20 hours (Figure 6 and S15). Analogous to the in vitro results, the probes had good S/B ratios (not shown) and remained stable in the medium for 20 hours. A mediocre signal was observed for the free dioxetane 2 only in the presence of Gram‐positive bacteria, while no chemiluminescence could be detected in Gram‐negative bacteria (Figure 6E,F and Figure S16). As mentioned in the introduction, the double‐layered cell wall of Gram‐negative bacteria acts as a tight barrier, preventing efficient dioxetane accumulation and activation. In contrast, all five siderophore conjugates displayed a strong activation by at least three out of the six tested ESKAPE bacteria.

Figure 5

Probe‐induced growth recovery in siderophore‐deficient E. coli and P. aeruginosa mutants. A) Growth recovery in the E. coli wildtype and enterobactin (ENT)—deficient strain ΔentA. The relative growth normalized to ENT is plotted in %. B) Growth recovery in the P. aeruginosa wildtype and pyoverdine, pyochelin (PYO)—deficient strain. The relative growth, normalized to PYO is plotted in %. All bacteria were grown in phosphate‐buffered LMR medium and incubated in the presence of 10 μM compound (or 1 % DMSO) and 10 μM FeCl3 for 48 hours at 37 °C, n=3. Error bars correspond to±standard error of mean (SEM).

Figure 6

Chemiluminescence kinetics in bacterial pathogens. A) Conjugate 8±Gram‐negative bacteria (E. coli, P. aeruginosa, K. pneumoniae, A. baumannii). B) Conjugate 8±Gram‐positive bacteria (E. faecium, S. aureus). C) Conjugate 9 ± Gram‐negative bacteria E. coli, P. aeruginosa, K. pneumoniae, A. baumannii). D) Conjugate 9 ± Gram‐positive bacteria (E. faecium, S. aureus). E) Control 2 ± Gram‐positive and ‐negative bacteria (E. faecium, S. aureus, E. coli, P. aeruginosa, K. pneumoniae, A. baumannii). Dotted lines correspond to the ± standard error of mean (SEM), (n=3). F) Total photon count of 2, 8 and 9 over 20 h. All experiments n = 3, final probe concentration of 10 μM in iron‐depleted, cation‐adjusted medium (IDCAM). The error bars correspond to the ± standard error of the mean (SEM). t max indicates the time point with the highest luminescence signal.

DOTAM probe 4 and MECAM probe 6 were activated by E. coli, A. baumannii, S. aureus, and a lower signal for P. aeruginosa and K. pneumoniae was observed (Figure S10A–B). Monocatechol probe 7 showed chemiluminescence for K. pneumoniae, A. baumannii and the two Gram‐positive strains. Commonly, the catechol probes reached their maximum intensity within 15–30 minutes after onset, then declined 103–105 fold within 5–7 hours. We assume that the time delay of the signal reflected a lag and delay phase of bacterial growth after their exposure to the IDCAM medium, while the following exponential growth led to enhanced chemiluminescence. The DFO dioxetanes 8 and 9 both showed a bright signal for all six bacteria, with a two‐ to four‐fold higher intensity for the Gram‐negative strains (Figure 6A–D). The TML trigger of probe 8 achieved higher luminescence intensities on average than probe 9 carrying a β‐galactosidase trigger. A similar complementation assay with the chromogenic substrate ortho‐nitrophenyl‐β‐D‐galacto‐pyranoside (OPNG) was conducted, as only a subset of the bacterial strains was reported to hydrolyze β‐glycosidic bonds. A shift from colorless to bright yellow (λ Ex=400 nm) due to the release of the ortho‐nitro phenol was observed for all strains, thus confirming an unanimous β‐galactosidase activity with broad substrate tolerance (Figure S17). Next, we verified the stability of 8 and 9 in the bacterial culture supernatant, in order to underpin their activation merely inside bacteria.

Bacteria are known to secrete enzymes into the surrounding to promote the extracellular decomposition of macromolecules into smaller products, than can then be taken up as nutrients. Therefore, an extracellular secretion and unspecific activation by microbial enzymes seems possible.[ , ] TML‐triggered 8 remained inactive and stable in the presence of all supernatants (Figure S18A,B). In contrast, 9 showed luminescence in K. pneumoniae supernatant, that was ≈20‐fold weaker compared to the incubation with the live pathogen. An OD600nm measurement after 20 hours excluded a bacterial contamination as the source for this lower probe activation (Figure S18E). No induction was observed by culture supernatants from the other ESKAPE pathogens (Figure S18C,D). Based on the superior stability properties in the culture supernatant, probe 8 was selected as a frontrunner within the set. We synthetized an analogous fluorescent probe 10 with umbelliferone as the fluorescent reporter (Figure 7A) to compare its efficiency in vitro and upon bacterial incubation with that of 8. In PBS, a high background signal was observed, and upon activation with NaBH4 the fluorescent signal rapidly reached its characteristic plateau (Figure 7A). The S/B ratio of 10 was 110‐fold lower than for the analogous chemiluminescent probe 8 and more than 40‐fold lower compared to dioxetane 2, which had the lowest ratio amongst all compounds (Figure 7B,C). The higher background of many biological samples in the blue‐cyan visible light range, where the umbelliferon of 10 is emissive, did not allow a direct comparison of the chemiluminescence and fluorescence outputs. However, also when directly comparing fluorescent emissions of 8 vs. 10 at 540 nm, we found that the fluorescent S/B ratio for 8 was 15‐fold higher compared to that of 10 (Figure S10). In bacteria, 10 was activated by four out of five tested ESKAPE pathogens, with a minor activation in E. coli, and reached its plateau after seven to ten hours. The proportion of average background signal for the fluorescent probe in the presence of Gram‐negative bacteria was more than twice as high (23.9±11.0 %) compared to the chemiluminescent counterparts 8 and 9 (8.5±6.6 %).

Figure 7

Characterization of turn‐on fluorescent TML‐coumarin DFO conjugate 10. A) Structure of 10. The siderophore is shown in black, the coumarin dye in green and the TML trigger in pink. B) Fluorescence kinetics after chemical activation of 10 (10 μM) ±1 mM NaBH4 in PBS at pH 7.4 over 2.5 hours. n = 3, dotted lines depict ± standard error of mean (SEM). C) Signal‐to‐background ratio of 10 ± 1 mM NaBH4. Error bars correspond to ± SEM. D) Fluorescence kinetic profiles of 10 (10 μM)±bacterial pathogens, n=4, dotted lines depict ± SEM. E) Summed fluorescence intensities, n=4, error ± SEM.

Overall, the comparisons demonstrate a clear superiority of chemiluminescent vs. fluorescent modes of detection. Both DFO probes showed a bright luminescence signal for all ESKAPE pathogens, including critical pathogens from the WHO priority list. As mentioned above, the prerequisite for a luminescence signal is the recognition and internalization through a chelator‐specific siderophore receptor, followed by subsequent activation. The broad scope of both probes are probably due to the wide prevalence of ferrioxamine siderophore transporters i.e. FhuD2 in S. aureus, FoxA in P. aeruginosa, FhuE in E. coli or A. baumannii,[ , ] as well as in K. pneumoniae. Upon cation‐DFO supplementation, previous studies revealed an increased pathogen growth, effects on biofilm formation and a modulated severity of infection. Taken together, these results indicate the acceptance of DFO as a xenosiderophore with multifaceted modes‐of‐action.[ , ] Additionally, a screening with gallium‐68‐complexed DFO in various strains confirmed the accumulation of the latter in a broad‐spectrum of clinical, prokaryotic pathogens. Some opportunistic pathogens can adapt smoothly to various environments and persist within cells of the host, e.g. in infections of the lung.[ , ] To demonstrate the applicability of the best probe 8 for the detection of intracellular pathogens, an infection model in confluent A549 human lung epithelial cells (LECs) was applied (Figure 8A). Cells were infected with the facultative intracellular pathogens P. aeruginosa PAO7 or S. aureus in the iron‐depleted medium at a multiplicity of infection (MOI, 9.52–9.92×105 CFUs) of 10. After 1.5 h, the extracellular bacteria were removed either by a rigorous washing step or treatment with gentamicin, that lacks the ability to penetrate into mammalian cells. Then the pre‐incubated siderophore probe (10 μM) was added to detect residual bacteria. Moreover, we monitored whether any probe was unspecifically activated by uninfected LECs. The presence of intracellularly residing bacteria after the infection was verified by lysis, plating und subsequent CFU counting of the serially diluted cell lysates (Figure 8E). In a previous study by Son et al., unconjugated 1 was readily activated in vitro by NQO1, an eukaryotic quinone oxidoreductase, which is overexpressed in A549 cells.

Figure 8

Detection of intracellular bacteria in A549 lung epithelial cells. A) Experimental workflow for A549 lung epithelial cell (LEC) infection and subsequent bacterial chemiluminescence imaging of P. aeruginosa or S. aureus at a multiplicity of infection (MOI) of 10. B) Chemiluminescence kinetic profiles of LECs infected with S. aureus followed by gentamicin treatment or a thorough wash and incubation with 8. C) Chemiluminescence kinetic profiles of LECs infected with P. aeruginosa followed by gentamicin treatment or a thorough wash and incubation with 8. D) Chemiluminescence kinetic profiles of LECs treated with gentamicin or with a thorough wash and incubated with 8. E) Quantification of intracellular bacteria after infection of A549 LECs with S. aureus or P. aeruginosa. F) Summed luminescence values for S. aureus treatments including controls. G) Summed luminescence values for P. aeruginosa treatments including controls. Dotted lines and error bars correspond to ± standard error of mean (SEM), n=3–6. All experiments in iron‐depleted, cation‐adjusted medium (IDCAM). The summed intensities in F) and G) were compared by two‐way ANOVA (****, p<0.0001). Gent.=gentamicin.

Indeed, incubation of 1 with sterile or bacteria‐infected LECs resulted in bright luminescence (Figure S19A). Thus, the unconjugated dioxetane was unable to distinguish eukaryotic cells from a prokaryotic infection. In contrast, 8, essentially a siderophore‐conjugated 1, showed a 10 000‐fold lower signal in the presence of sterile LECs. However, bacteria‐infected LECs showed bright activation and had five‐ to eight‐fold higher luminescence intensities compared to baseline (Figure 8B,D). This difference became even more apparent when comparing the respective summed luminescence values (Figure 8F,G). Similar correlations between the treatment groups were found for a fluorescent readout of the weak benzoate fluorophore IV (λ Ex=340 nm), which was formed following chemiexcitation (Figure S20 and Figure 1), while the kinetics with an increasing emission reflected the accumulation of the fluorophore over time. Two further experimental groups were studied. The permanent addition of gentamicin to the RPMI medium restricted extracellular bacterial growth even after the probe addition; this led to a 2–3 fold reductions of chemiluminescence compared to gentamicin‐free incubations (Figure 8B,C and F,G).

Secondly, bacterial transporters were blocked with a tenfold excess of free DFO before and during infection; this led to 7–11 fold reductions of chemiluminescence compared to DFO‐free incubations (Figure 8B,C and F,G). These findings underline the siderophore's key role for an efficient translocation and activation via the bacterial iron transport systems. In sum, all events that yielded a bright probe activation were infection‐ and siderophore‐dependent, and thus demonstrated the probe's ability to distinguish even small amounts of bacteria from sterile host cells. A comparable curve shape and similar intensity was seen for the DOTAM probe 4, while monocatechol 7 reached even higher luminosity values (Figure S19). In either case, an infection was required to obtain a signal. Previous studies reported that the free DFO siderophore accumulated randomly by fluid‐phase endocytosis into mammalian cells. This process was assumed to be much slower than the active transport by bacterial siderophore transporters. The induction of luminescence by the bacteria alone took several hours (Figure 6), and an even slower onset of signals would be expected for intracellular bacteria within LECs.

As the opposite, i.e. an instant luminescence signal was observed, we hypothesize that the probes were not activated by bacteria within LECs, but by small amounts of released enzymes. These may stem from the bacteria, or also from the LECs, as bacterial infections have been reported to cause membrane damage and a spill of NQO1 from the mammalian cells’ cytosol. Taken together, this renders siderophore dioxetane conjugates promising candidates for the in vitro detection of bacteria, but also for the sensitive monitoring of bacterial infections in tissues. For a potential use of the probes as a diagnostic tool to detect bacteria, the ability to detect only small numbers of colony forming units (CFU), such as usually found in medical samples, is important. To this end, the limits of detection (LODs) of live S. aureus and P. aeruginosa bacteria by chemiluminescent 8 and its fluorescent analog 10 were determined under iron‐depleted conditions. Optical density (OD600) measurements were linked to numbers of living bacteria by colony forming unit (CFU) counting through plating after iron starvation (see the Supporting Information). S. aureus or P. aeruginosa bacteria were starved for ferric iron and subsequently serially diluted in IDCAM prior to incubation with 10 μM probe (Figure 9A). After 24 hours of luminescence (Figure 9B/C) or fluorescence (Figure 9D/E) recording, the signals were integrated and plotted against the previously obtained bacterial count using logarithmic scales. Best fit lines showed a linear decline of the signal upon reduction of the initial CFUs/mL for all tested strains and probes. Chemiluminescent 8 displayed a later decline of the summed signal at lower bacterial concentrations than observed for the fluorescent probe 10, and hence could detect a smaller number of Gram‐positive or ‐negative pathogens.

Figure 9

LOD determination of siderophores probes. A) To determine the limit of detection (LOD), cultures of S. aureus or P. aeruginosa were iron‐starved, then serially diluted in iron‐depleted, cation adjusted medium and then incubated with chemiluminescent probe 8 or fluorescent probe 10 (10 μM each). Total integrated signal after 24 h of incubation of B) 8 with S. aureus, C) 8 with P. aeruginosa, D) 10 with S. aureus and E) 10 with P. aeruginosa (mean ± SEM, n=4) is shown. Horizontal lines show the mean signal of probe in medium ± SEM (dashed) of control samples lacking bacteria. Best‐fit lines show linear regressions of the log‐transformed data. For all experiments, each sample was compared to the no‐bacteria control by one‐way ANOVA (***, p=0.001–0.005, **** p<0.0001). The signal‐to‐background (S/B) ratios were plotted against different S. aureus F) and P. aeruginosa G) concentrations for the two probes. The S/B ratios of the two probes were compared by two‐way ANOVA (****, p<0.0001).

From the summed values, the S/B ratios for each probe could be calculated and were plotted against the logarithmic bacterial concentration (Figure 9F/G). Remarkably, 8 exhibited a LOD value of 9.1×103 CFU mL−1 for S. aureus, while 10 could minimally detect 2.8×106 CFU mL−1 of the same pathogen. Similarly, 8 exhibited an LOD value of 5.0×104 CFU ml−1 for P. aeruginosa, while 10 reliably detected 1.0×107 CFU mL−1. The strongly enhanced sensitivity of 8 (307‐ and 200‐fold, respectively) demonstrates a clear sensitivity advantage of chemiluminescent siderophore dioxetane probes for the detection of bacterial pathogens compared to fluorescent analogs.

The ability to detect bacterial pathogens at low abundance in a clinical setting was probed by spiking S. aureus and P. aeruginosa at their LOD into sterile, human plasma, followed by addition of 8. The luminescence (and also the fluorescence) signals were significantly higher in the presence of both prokaryotes compared to 8 in sterile plasma (Figure S21). Thus, the probe displayed sufficient stability in a complex biological environment to enable bacterial pathogen detection.

Previous microbiological studies demonstrated that the bacterial CFU counts recovered from clinical samples (e.g. blood of patients) were commonly low. PCR‐based methods have shown lower LODs (30–700 CFU mL−1), but fail to distinguish DNA from live vs. dead bacteria or debris. Microfluidic methods, which employ antibody‐coated microspheres and subsequent optical analysis to identify bacteria generally, had LOD values in a similar range (103–105 CFU mL−1) as 8. A commercial bacterial viability kit (BacLight®) with fluorescent reagents exhibited a LOD with more than 106 CFU mL−1 of bacteria. Compared to previous chemiluminescent probes for the detection of Salmonella, Listeria and Mycobacteria, 8 detected a similar bacterial concentration range (103 to 104 CFU mL−1).[ , ]

Conclusion

In this study, we expanded the application of artificial siderophores as functional and versatile targeting entities to the chemiluminescent imaging of bacterial infections. DOTAM‐ and MECAM‐based siderophore mimics were conjugated to enzyme‐triggered dioxetanes in up to nine synthetic steps. All probes showed bright luminescence and very high S/B ratios in vitro, but hydroxamates were superior to catecholates due to the quenching effects of the latter. Notably, five siderophore conjugates retained their ability to shuttle iron into the bacterial cell, thereby enabling the reliable detection of a broad range of Gram‐negative and Gram‐positive pathogens. The advantages of a chemiluminescent vs. fluorescent detection principle of bacterial pathogens manifested in higher S/B ratios in vitro and in cells, that translated in lower limits of detection. Moreover, 4, 7 and 8 succeeded to detect facultative intracellular pathogens in bacteria‐infected LECs, and remained inactive when incubated with uninfected host cells.

The study suggests that siderophore dioxetane conjugates may find valuable applications for the detection of microbes in food quality control as well as in the healthcare sector. It is a good basis to examine trigger‐siderophore combinations more systematically and optimize them for high conjugate enrichment. In addition, the probe's performance across a broader panel of clinically relevant, drug‐sensitive and drug‐resistant clinical isolates per strain needs to be investigated further. Future projects will also include the development of IR‐shifted dioxetane siderophore conjugates, which permit the broad‐spectrum, chemiluminescent in vivo imaging of bacteria in rodent infection models. Beyond practical perspectives, the study illustrates how a molecular targeting can significantly improve the sensitivity and specificity of drug delivery. It also underlines the attractiveness and potential of the bacterial siderophore uptake system as an entry gate for the transport of cargo.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.