Optimization of artificial siderophores as 68 Ga-complexed PET tracers for in vivo imaging of bacterial infections

The diagnosis of bacterial infections at deep body sites benefits from noninvasive imaging of molecular probes that can be traced by positron emission tomography (PET). We specifically labeled bacteria by targeting their iron transport system with artificial siderophores. The cyclen-based probes contain different binding sites for iron and the PET nuclide gallium-68. A panel of 11 siderophores with different iron coordination numbers and geometries was synthesized in up to eight steps, and candidates with the best siderophore potential were selected by a growth recovery assay. The probes [ 68 Ga] 7 and [ 68 Ga] 15 were found to be suited for PET imaging based on their radiochemical yield, radiochemical purity, and complex stability in vitro and in vivo. Both showed significant uptake in mice infected with E.coli and were able to discern infection from LPS-triggered, sterile-inflammation. The study qualifies cyclen-based artificial siderophores as readily accessible scaffolds for the in vivo imaging of bacteria.


INTRODUCTION
Infections with pathogenic bacteria are a cause of high morbidity and mortality and therefore constitute a major threat for human health. 1 This situation is exacerbated by the rise of antimicrobial resistance, rendering established treatments ineffective, while the pipeline of novel, resistance-breaking antibiotics remains thin. 2,3Today's diagnosis of bacterial infections is based on clinical symptoms and the analysis of biofluids, typically blood or urine, using microbiological, genetic and mass spectrometric techniques.However, the analysis of biofluids struggles to detect early-stage infections at deep body sites (e.g.heart, brain, or medical implants) that are hardly accessible for sampling.For such cases, noninvasive imaging techniques with molecular probes that localize bacterial infections bear the potential to improve diagnostic capabilities significantly, as outlined by recent reviews in this and other journals. 4,5,6In fact, infection imaging of vulnerable patient populations (e.g.following cancer chemotherapy or organ transplantation) has already become clinical practice, and it is mostly based on the detection of [ 18 F]fluorodeoxyglucose (FDG) by positron emission tomography (PET).Because FDG is taken up by all metabolically active cells, the tracer has low specificity and cannot distinguish between sterile inflammation and infection. 7,8erefore, some alternative approaches to identify bacteria-specific PET tracers have been pursued recently.For example, white blood cells (WBC) or antibodies could be labelled with 67 Ga, 99m Tc or 111 In and have been successfully applied in the imaging of osteomyelitis or prosthetic joint infections. 91][12][13][14][15][16][17][18][19] Also, essential bacterial metabolites have been converted to radioactive imaging probes. 202][23][24] Recently, D-[5- 11 C]Glutamine could specifically discriminate live E. coli and MRSA in dual-infection murine myositis model versus sterile inflammation caused by heatkilled bacteria. 25third strategy, pursued in this study, is to utilize the active iron transport systems of bacteria for a selective and pronounced uptake of the probe.To cover their continued demand for iron ions, bacteria synthetize small molecular weight iron-chelators, so-called siderophores (greek: sidero = iron, phoros = carrier), that are secreted to the environment and actively internalized by bacterial outer membrane receptors, once loaded with iron. 26cause Ga 3+ ions exhibit similar coordination properties as Fe 3+ ions, siderophore-based imaging probes have been proposed, which are loaded with the positron emitter gallium-68 instead of iron. 4 First in vivo studies have been conducted to image the fungus Aspergillus fumigatus with [ 68 Ga]triacetyl fusarine and ferrioxamine derivatives, 27,28 or Pseudomonas aeruginosa with gallium-68 complexed pyoverdine. 29Lately, a 68 Ga-labelled version of the clinically used antidote desferrioxamine (Desferal®, DFO-B) has been repurposed in an acute murine myositis model to image Gram-positive and Gram-negative bacteria in vivo. 30 note, siderophores have been widely employed as Trojan Horses, transporting a broad range of payloads efficiently into bacterial cells. 31The success of this concept has been underlined by the approval of the first siderophore-coupled antibiotic cefiderocol (Fetroja ® ) in 2019. 32While the above mentioned studies use the iron binding site of a natural siderophore to incorporate the PET-tracer, we envisaged to (i) design bifunctional compounds with two separate binding sites for iron and the radionuclide and to (ii) employ a non-natural siderophore analog that is potentially accepted by a broad range of bacteria.This decoupling of tracer binding and uptake allows to optimize both functions independently.In a previous study, we have qualified artificial siderophores based on the cyclen scaffold as suited to accommodate two metals, label a broad range of bacteria, and detect infections in mice using optical imaging with tagged fluorophores. 33While our previous probes were confined to cellular and small animal imaging, the current study reports a crucial step towards translation of the probes, i.e. their optimization to PET tracers for the imaging of large species including humans.

Synthesis of cyclen-based siderophore analogues
We aimed to use the versatile cyclen scaffold to attach catecholate units for Fe 3+ chelation and at the same time complex a Ga 3+ ion via the cyclen core itself as a PET imaging tracer.In our original design, 33 cyclen was functionalized to a tetrapodal 1,4,7,10tetraazacyclododecane-1,4,7,10-tetraacetic amide (DOTAM) to accommodate the metal for imaging. 34However, the reaction times for Ga 3+ incorporation were in the range of hours, which is incompatible with radiochemical synthesis and positron emitter half-life.Because the rates of metal incorporation are much higher for 1,4,7,10-tetraazacyclododecane-1,4,7,10tetraacetic acid (DOTA)-based cores with free carboxylic acid moieties, 35 we systematically varied the number of amide-linked catechols (for Fe-complexation) vs. number of carboxylates (for Ga-complexation) from 1:3 to 2:2 to 3:1 (Figure 1).In addition, the use of one DOTA and one DOTAM core that are conjugated to each other would enable a 3:3 constellation, i.e. a scaffold with three gallium-68 complexing carboxylates is linked to a scaffold with three iron-complexing catechol units.In order to facilitate Ga-complex formation from DOTAM moieties, α-methyl substituents, which should exert a Thorpe-Ingold effect and reduce the bite angle at the cyclen core, were introduced as well. 36The catechol units, representing bidentate iron binders, were partly masked as acetylated prodrugs, which are activated in situ in cellular environment, in order to prevent permanent inactivation of the catechol by enzymatic alkylation and to facilitate 68 Ga-complex formation. 37Starting from cyclen, we made use of optimized reaction conditions, which can yield each a specific substitution pattern, from a mono-to a tri-substituted core respectively (Figure 2).The metal-free ligands -referred to as 'precursors' -were obtained by slow, sequential addition of the appropriate equivalents of alkyl bromide and base in acetonitrile with near to complete conversion and high purity.Depending on the choice of the initial substitution reaction, this strategy led to 1:3 or 3:1 catechol:carboxylate ratios.The synthesis of mono-catechol siderophore analogues started with the reaction of cyclen and tert-butyl bromoacetate to yield 3 (Figure 2A).The tri-substituted cyclen 3 was used as a crude product and reacted with the linker 1 under basic conditions to attach the fourth arm in 55% yield over two steps.Subsequently free amine 5 could be afforded after hydrogenolysis of the Cbz protecting group with Pd/C under H2 atmosphere in 92% yield.The primary amine of 5 was reacted with the in situ-generated acid chloride of 2 in a two-phase Schotten-Baumann reaction to the desired amide 6 in 30% yield, which was subsequently deprotected with 50% TFA/AcOH to give the acetylated mono-catechol 7. The free catechol 8 was then obtained by transacetylation with 20% DIPEA in MeOH.
In order to obtain tris-catechol siderophores, cyclen was reacted first with the bromoacetamide 17 and subsequently with benzyl 2-bromoacetate under basic conditions, followed by a Boc group removal/hydrogenation sequence (Figure 2B).In this manner, intermediate 9 was obtained within four steps.The amine 9 was reacted with the freshly prepared acid chloride of 2 in a Schotten-Baumann reaction to give the acetylated precursor 10 in 32% isolated yield within six synthetic steps.A similar sequence with 9 and the acid chloride of 2,3-bis(benzyloxy)benzoic acid, followed by hydrogenolysis, yielded precursor 11 with free catechols.This compound could also be obtained by transacetylation of 10 with 20% DIPEA in MeOH in 31% overall yield within seven synthetic steps.The two cis-and trans-substituted di-catecholates 34 and 35 were obtained as an inseparable mixture by similar reaction sequences; the same was true for the regioisomers 25 and 26 that were alpha-methylated at the free carboxylic acids (Supporting Information, Figure S1 and Figure S2).Synthesis of siderophore-based PET imaging agents.(A) Synthesis of precursors (Figure 3B, Figure S3 and Figure S4).According to previous studies, the metal-DOTA complex formation is a two-step process. 38Initially, a reversible adduct (so-called type I complex) forms.Protonation/deprotonation reactions at the carboxylic acid groups and macrocycle nitrogen protons then result in the slow conversion to a highly stable type II complex, that is featured by a full coordination of the metal. 39,40 he rate of metal-DOTA complexation is higher at elevated temperature or with more flexible macrocycles, because the exchange rate between conformational isomers increases. 41,35 eprotonation associated with gallium complexation leads to two evident effects in the 1 H-NMR spectra: 1) higher chemical shift values due to a deshielding effect for the methylene 1 Hs close in space to the central Ga cation, and 2) a 1 H-signal split related to a fixation of the apparently fluxional DOTA unit into a rigid or semi-rigid complex.The affected protons become magnetically nonequivalent because of a defined complexation status, 39 which leads to the resolved H-NMR peak pattern shown in Figure 3B, Figure S3 and Figure S4.

Growth recovery capabilities of cyclen-based siderophore analogues
Next, we tested whether the artificial siderophores were able to enter bacterial cells in free as well as in Ga 3+ -complexed form.For this purpose, a surrogate assay that measured the compound-mediated delivery of iron into bacteria based on growth recovery was applied (Figure 4A). 33The ΔentA and ΔentB mutants of E. coli BW25113, that cannot biosynthesize their endogenous siderophore enterobactin (ENT), were unable to grow under iron-depleted conditions as well as under Fe 3+ supplemented conditions (Figure 4B and Figure S5A, DMSO control).Because these strains are reliant on exogenous siderophores to grow under ironlimited conditions, the exogenous addition of the natural siderophore ENT (10 µM) restored the growth of the E. coli ΔentA and ΔentB mutants.Remarkably, also the supplementation of DOTAM-based siderophore analogues (10 µM) fully rescued the growth in seven out of eleven cases for both mutant strains.The most efficient growth recovery was observed for the DOTA tris-catechol 15, as well as the tris-catechol 10 and 11.Also the di-catechols 25/26 and 34/35 were able to relieve iron deficiency.In contrast, the mono-catechols were generally less potent.Interestingly, also [ nat Ga]11 was accepted as a siderophore and exhibited a potency that was comparable to the uncomplexed analog 11.A less pronounced, but still measurable growth recovery can also be found for the mono-catechols 7 and [ 69 Ga]7.
The decetylated mono-catechol 8 was found to be more efficient than 7 in the E. coli ΔentA mutant.The effect of α-methyl substituents can be assessed by comparing the methylated mono-catechol 8 with its unmethylated congener 24.Likewise, the methylated tris-catechol 37 can be compared to the unmethylated 10.In both cases, the methyl-substitution led to an impairment of growth recovery, indicating a decreased uptake of the complexes.In sum, these results demonstrate that the cyclen-based probes can shuttle Fe 3+ into the bacterial cell, indicating their internalization, even when they are loaded with a Ga 3+ cation.
Growth recovery data from the pyochelin and pyoverdine-deficient P. aeruginosa strain ΔpvdD ΔpchE-F showed that the exogenous addition of DOTAM-based siderophore analogues could relieve iron deficiency to an extent that was comparable to that of the natural siderophore pyoverdine (Figure S5.B).A significant increase in efficiency compared to its result in the E. coli mutants was observed for compound 7. Compounds 8, 10 and 15 were found to exhibit retained activity in the P. aeruginosa mutant.The similar efficiency across two bacterial species is encouraging for the applicability of our probes.
With a read-out after 48 hours, the growth recovery assay mimics the long-term iron supply into bacteria.An in vivo PET scan with a gallium-68 (t1/2 = 67.71min 42 ) complexed tracer takes about 60 minutes and is rather short in comparison.In vitro uptake experiments during similar timeframes with radiolabelled siderophore analogues would resemble the in vivo setup more closely, and an estimation of the maximum and minimum tracer uptake could be gained by comparing wildtype and siderophore-deficient strains.Those studies will be important for a future candidate selection.
According to the literature, the release mechanism of ferric iron from siderophores is siderophore-and species-dependent. 43Release can occur in the periplasm or cytoplasm, and it is mediated either by intracellular Fe-siderophore reductases or Fe-siderophore hydrolases. 44,45For example, the tris-catechol siderophore enterobactin is degraded via a lactone cleavage by esterases like Fes or IroD in the cytoplasm, followed by the reduction of Fe-(III) to Fe-(II). 46The mechanism for artificial DOTAM siderophores is unknown.Given that they lack a hydrolysable backbone, we assume that the release is either mediated by metal reduction, or by the metabolism of the catechol units.The mono-and di-catechol siderophores with a weaker binding affinities and redox potentials are easier amenable to ligand exchanges as well as reductions.

Chemical stability and cytotoxicity
The chemical stability of the Ga 3+ -free precursors 7, 8, 24, 11, 37, and 34/35 was measured in PBS at 37 °C over a period from 24 to 96 hours by LC/UV/MS (see Figure S6).
All compounds showed a stability of at least 60% over 4 hours under the given conditions, which is deemed as sufficient for PET measurements (Figure S6A).For prodrugs, the acetyl group stability gradually declined, and no acetylated compound could be detected after 24 hours (Figure S6B), which is in accordance with previously published data. 33The cellular cytotoxicity of Ga 3+ -free precursors 7, 8, 11, 15, 24, 25/26, 34/35, and 37 as well as Ga 3+complexed siderophores [ nat Ga]7 and [ nat Ga]10 was tested in the five eukaryotic cell lines L929, A549, KB-3-1, MCF-7 and FS4-LTM.All compounds were classified as non-cytotoxic up to concentrations of 100 M (Table ST1).

Radiochemical synthesis
Representative mono-, di-and tris-catecholates (7, 25/26, 34/35, 10, 11, 15), that induced growth recovery, exhibited good stability in PBS and showed no cytotoxic effects, were selected for radiochemical labeling with gallium-68.For manual labelling, 30 -50 nmol precursor were mixed with unprocessed gallium-68 chloride.The complexation process was assessed by (i) radiochemical yield of the crude product determined from an aliquot of the reaction solution by HPLC (RCY (crude)), (ii) the radiochemical yield (RCY) of the purified tracer, defined as the amount of activity (decay corrected) in the product expressed as the percentage (%) of related starting activity, and (iii) the radiochemical purity (RCP), defined as the proportion of the total radioactivity in the sample which is present as the desired radiolabeled species. 47The precursor 15 reached the highest values for RCY (crude) (97.8%),RCP (84.3%) and RCY (56.7%), followed by the mono-catechol 7 and the triscatechol 10 (Figure 5A).Interestingly, 11 demonstrated dramatically decreased RCY (crude) and RCP compared to the acetylated prodrug 10.Likewise, the complexation of 8 with 'cold' gallium ions was less efficient than that of 7 (data not shown).Stable complexes with enterobactin mimics such as TREN-CAM (tripodal enterobactin analogue-catecholate coordinating moieties), which resisted transchelation even with 1000-fold excess of desferrioxamine, provide evidence for the gallium-68 complexing ability of catechols. 48In contrast to these compounds, our siderophore analogues exhibit two chelation centers -at the catechol units and at the cyclen moiety.Therefore, we introduced transient masking groups (acetyl) at the catechols, to facilitate the rapid formation of stable gallium-68 complexes at the cyclen unit.The significantly higher RCY for [ 68 Ga]10 compared to the deacetylated tracer [ 68 Ga]11 support this assumption.Because a fast complexation is imperative due to the short half-life of gallium-68, the precursor 11 was deselected.In contrast, 15 with a spatially separated DOTA moiety achieved high RCY (crude), RCP and RCY without acetylated groups.The di-catecholates 25/26 and 34/35 had rather low RCPs and RCYs.Because 25/26 and 34/35 had the additional drawback of being regioisomeric mixtures, the di-catecholates were no longer pursued.Thus, we conclude that the spatial separation of an optimal 68 Ga-chelator (DOTA) from the iron-chelating DOTAM siderophore facilitates the formation of stable 68 Ga-complexes.Based on the radiochemical data, the three siderophores 7, 10 and 15 were selected for further studies.
The radiochemical synthesis of [ 68 Ga]7, [ 68 Ga]10 and [ 68 Ga]15 was first optimized manually regarding e.g.EtOH content, heating time, pH value and purification cartridges (Table ST2 in the Supporting Information).Based on the manual procedure, the precursors were complexed on an automated synthesis module to minimize radioactive exposure, standardize the complexation conditions and corresponding RCP/RCY for in vivo studies (Figure 5B).30 nmol of 7 could be labelled within 8 minutes and purified via a t C18 light cartridge, while precursor 15 was labelled within 6 minutes and purified via a HLB light cartridge.Precursor 10 needed 25 minutes of synthesis time.The longer time might be attributed to the presence of only one free carboxylate in the cyclen and increased sterical hindrance by the side groups.Ga-complexed tracers was determined in an octanol/water matrix, with [ 68 Ga]7 being the most hydrophilic (logD = -3.3)and [ 68 Ga]15 notably more lipophilic (log D = -1,6, Figure 5C).
The stability of the radiotracers in PBS and in human serum at 37 °C was evaluated by radio chromatograms directly after synthesis for two hours (Figure 5D-F and Figure S7).As found for the uncomplexed precursors, all three compounds were >80% stable in PBS for two hours.However, [ 68 Ga]7 and [ 68 Ga]10 displayed some instability in serum: For [ 68 Ga]7, 19.1 ± 6.2% gallium-68 were released after two hours of incubation.Moreover, the tracer deacetylated, resulting in 58.9 ± 22.0% deacetylated [ 68 Ga]7 within 2 hours.91.4 ± 1.8% of [ 68 Ga]10 were deacetylated in serum after two hours.Stepwise enzymatic deacetylation in human serum, and hence release of the free catechol moieties of both complexes over the course of two hours, was accompanied by a shift to smaller (more polar) retention times (see Figure S7A and C right), corresponding to the conversion of the prodrug to the active drug.
As [ 68 Ga]15 did not bear any acetyl groups, deacetylated metabolites were not observed, but a more lipophilic metabolite was formed after 2 hours (see Figure S7B).Still, 80.8 ± 0.8% [ 68 Ga]15 remained stable for more than two hours in serum.While [ 68 Ga]15 displayed high RCY and tracer stability compared to other bacteriaspecific PET tracers, 12,18,25,27,29,30 [ 68 Ga]7 could be further improved by either installing a fourth carboxylate donor group, or instead a smaller NOTA chelator to form more stable gallium-68 complexes. 49Free catechol-assisted trans-chelation or hydrolysis was successfully masked by the introduction of transient protection groups.Deacetylation of the catechol moieties occurred over ca.6 hours in PBS (pH 7.4), as shown in Figure S6B and in a previous study. 33,50 owever, stability and radio HPLC data of [ 68 Ga]7 (Figure 5 and Figure S7), suggest a more rapid reaction to the free catechol in blood than in PBS.Therefore alternative masking groups with higher stability, e.g.sialylethers or acetals, could be evaluated with respect to species-specific higher hydrolase activities. 51
[ 68 Ga]7 exhibited almost exclusively renal clearance with low blood pool retention and fast clearance from examined organs (Figure 6A).In the last frame, unspecific tracer uptake in the healthy muscle tissue was negligible (0.05 ± 0.01 %ID/g, Figure 6B).[ 68 Ga]15 showed a similar kinetic compared to [ 68 Ga]7, i.e. a rapid, mainly renal clearance and a low blood pool retention, but exhibited a higher liver accumulation (Figure 6D), correlating well with its higher lipophilicity.The uptake into healthy muscle amounted to 0.34 ± 0.05 %ID/g (Figure 6E).[ 68 Ga]10 was excreted via the kidneys and liver, but showed a higher background activity in blood, with longer average retention periods in the respective imaged organs (Figure S8).The non-specific uptake of [ 68 Ga]10 in healthy muscle during the last 10 minutes of imaging amounted to 0.5 ± 0.09%ID/g.Only 36.6 ±11.5% of [ 68 Ga]7 could be detected in blood after one hour of imaging, as 63.4% ± 11.5% hydrophilic metabolites were formed instead.In urine the tracer was mainly detected in its deacetylated, active form (ca. 75.4%), with a highly hydrophilic metabolite amounting to 2.0 ± 0.9% and the acetylated, intact [ 68 Ga]7 to 22.6 ± 11.7% (Figure 6C).
Stable [ 68 Ga]15 was detected in blood as well as in urine, while the formation of metabolites was higher in blood (14.6%) than in urine (1.6%)Only 1.3 ± 1.5% acetylated and mostly deacetylated (95.4 ± 3.5%) [ 68 Ga]10 could be detected in blood.The tracer was detected in urine in a relative quantity of 42.1% ± 35.1%, accompanied by formation of metabolites that accounted for 29.8 ± 25.1% (Figure S8).Moreover, [ 68 Ga]10 seems to exhibit a loss of gallium-68, since an increased liver uptake and the highest unspecific uptake in healthy muscle was observed, which indicates a transmetalation with transferrin in blood. 52Based on the poor radiochemical yield and unspecific blood pool retention, [ 68 Ga]10 was not considered for the subsequent infection model.

PET/CT imaging of E. coli infected muscle vs. sterile-inflamed muscle in mice
The best suited tracers [ 68 Ga]7 and [ 68 Ga]15 were tested in an in vivo infection model.
In order to evaluate whether the tracers could differentiate between a bacterial infection and a sterile inflammation, 3 x 10 7 CFU's of E. coli were injected into the Musculus gastrocnemius of the left leg, and 27 µg lipopolysaccharide (LPS) were injected into the right leg in male C57BL/6N mice.LPS induced a sterile inflammatory reaction, which should not be detectable by tracers that specifically visualize bacteria.Twenty hours after this challenge, 11.6 ± 0.8 MBq (0.31 ± 0.02 mCi) [ 68 Ga]15 (n =8) or 11.8 ± 1.5 MBq (0.32 ± 0.04 mCi) [ 68 Ga]7 (n =6) were injected i.v., respectively, and dynamic PET scans were acquired.
[ 68 Ga]7 showed a rapid clearance in the infection setting, as already observed in uninfected mice.A higher uptake of the tracer into the substantially enlarged infected muscle compared to the inflamed muscle was evident by visual inspection of the PET scan (Figure 7A and Supplementary Video SV1).The tracer accumulated rapidly in both legs within the first 10 minutes, but the subsequent decrease was faster in the sterile-inflamed muscle (Figure 7B).Semi-quantitative analysis, based on the sum of regions of interest (ROIs) of the last half hour of dynamic PET scan, showed an increased uptake of [ 68 Ga]7 into the infected muscle of 0.67 ± 0.05 %ID/g in comparison to the sterile-inflamed muscle (0.36 ± 0.06 %ID/g uptake, p < 0.0001; Figure 7D).The accumulated dose %ID/g is 11-fold higher in the infected muscle than in the healthy muscle (see Figure 6).This indicates that [ 68 Ga]7, despite its structural simplicity, can distinguish between bacterial infections, sterile inflammation and healthy tissue in vivo.The autoradiographic examination of the dissected muscles confirmed the PET scan results, as an increased uptake of the tracer into the infected vs. the inflamed muscle was clearly visible (Figure 7C).Furthermore, Gram staining of the dissected muscles confirmed the presence of bacteria in the infected muscle, whereas expectedly, no bacteria were detected in the sterile LPS inflammation.On the contrary, immune staining with an antiCD38 antibody identified invasion of macrophages in both muscles (Figure 7C).
For the DOTA tris-catechol [ 68 Ga]15, tracer clearance in E. coli / LPS-challenged mice was as rapid as in non-infected animals.Similar to [ 68 Ga]7, an increased uptake into the infected muscle was visible 15 minutes after injection (Figure 8 and Supplementary Video SV2).A semi-quantitative analysis of the last frame revealed that the uptake into the E. coliinfected muscle amounted to 0.51 ± 0.18 %ID/g, 1.7-fold higher than into the sterile-inflamed muscle (0.30 ± 0.09 %ID/g; p = 0.035; Figure 8D).The autoradiographic imaging of the dissected muscle also confirmed a higher uptake of [ 68 Ga]15 into the E. coli-infected muscle (Figure 8C).In comparison to [ 68 Ga]7, a lower %ID/g over time could be accumulated, but a differentiation between infected and sterile-inflamed muscle was possible.Autoradiographic images of the dissected muscles confirmed the PET/CT results, and even showed a stronger difference between the infected muscle and the muscle with sterile LPS inflammation.Gram staining confirmed the presence of bacteria in the infected muscle, while no bacteria were detected in the sterile LPS inflammation.Expectedly, immune staining with an anti-CD38 antibody identified invasion of macrophages in both muscles.The CD38-positive stains (Figures 7 and 8) figuratively depict the common response of the immune system (macrophage infiltration), to E. coli infection and sterile inflammation.In general, the bioavailability of free iron lies in the order of ~10 −18 M, i.e. very low.Upon LPSinduced sterile inflammation, an iron anemic environment is observed at the injection site, with low Hb, even lower iron content in Hb and elevated levels of hepcidin. 53An enrichment of siderophore analogues at the inflammation site in the absence of bacteria would have to occur through unspecific host or immune cell uptake, but this was not indicated in the biodistribution studies (Figure 6).Moreover, the TonB-dependent siderophore uptake machinery is unique to prokaryotes, and active internalization by bacterial siderophore receptors occurs faster than unspecific uptake by host cells.Thus, the hypoferremia of nonsterile inflammation as well as the probe uptake mechanism all constitute advantages that contribute to the specificity of siderophore-based bacterial imaging probes.
An important point concerns the pathogen specificity and selectivity of the probes.
Prokaryotes are able to hijack iron-bound siderophores synthetized by other pathogens as so-called "xenosiderophores".An invading microbe expresses a manifold of heterologous and homologous, chelator-specific siderophore outer membrane receptors (OMRs) for the acquisition of iron from the most readily attainable source. 54However, PET imaging with natural siderophores like pyoverdine produced by P. aeruginosa was highly specific for this bacterial species and showed no accumulation in others. 29The uptake of a gallium-68labelled DFO-B tracer via hydroxamate OMRs (e.g.FpvA) into P. aeruginosa and S. aureus was successful as expected, but marginal in E. coli strains lacking the DFO-B transport machinery. 30The siderophore analogues described herein target catechol-specific OMRs, known to be expressed for example in P. aeruginosa (PfeA), 55 E. coli and K. pneumoniae (FepA) 56,57 .Further studies indicate that other pathogens like S. aureus, S. thyphimurium and Y. enterocolitica are also able to seize catechol-based xenosiderophores like ENT for their iron supply. 58The catecholate-based siderophores anchored on an artificial DOTAM scaffold was shown to label to a broad range of pathogens before. 33This demonstrates its function as a xenosiderophore, and we hypothesize that simple artificial scaffolds may find broader acceptance compared to distinct natural siderophores like pyoverdine.However, the specificity of the current probes [ 68 Ga]15 and [ 68 Ga]7, and also the variance among isolates of a given species including defined, catecholate receptor-deficient strains, requires an experimental determination in future studies.
While the cyclen scaffold allows the versatile attachment of siderophore groups as well as additional effectors, it has disadvantages concerning tracer accommodation compared to NOTA based chelators.0][61] Due to the reduced complex stability, cyclen scaffolds run the risk of transmetalation leading to decreased RCY.Second, DOTA chelators exhibit a slower complexation kinetics and accordingly require a higher temperature and elongated reaction times compared to NOTA complexes. 62In line with this, we could observe that more substitutes on the cyclen scaffold lead to a slower complex formation, reflected in the gained radiochemical yield and the elongated reaction time.Thus, an expansion of the concept to NOTA chelators might lead to further improved probes.

CONCLUSION
We have expanded the suite of applications for artificial, cyclen-based siderophores to PET imaging, in order to enable the monitoring of infections in large animals.The compounds were synthetized via robust and scalable synthetic pathways in up to eight steps in a modular manner and discriminated through a suite of chemical and biological assays (Figure S11).
Notably, the compounds could simultaneously host a 68 Ga 3+ for PET imaging in their cyclen core and still serve as efficient xenosiderophores in E. coli.A reduced number of catechols clearly led to lower growth recovery.However, two and even one catechol are capable of a growth recovery; this finding is in line with the fact that many siderophore-antibiotic conjugates employ a single bidentate iron-binding moiety to enhance their uptake. 31The αmethyl variants did neither improve bacterial growth recovery nor radiochemical labeling, and therefore seems redundant in the scaffolds investigated here.For six selected precursors, manual and automated radiochemical synthesis procedures were established successfully to yield the respective 68 Ga-complexed PET tracers.Two bacteria-targeted PET tracers, [ 68 Ga]7 and [ 68 Ga]15, displayed favorable biodistribution and stability properties and could reliably distinguish E. coli infections from LPS-induced, sterile inflammation in mice.While active bacterial uptake should be superior in the DOTA tris-catechol [ 68 Ga]15, the ease of synthesis but also the imaging data speak for [ 68 Ga]7.These compounds expand the yet limited arsenal of molecular probes for bacterial imaging in large animals.In particular, we found that addressing the bacterial iron uptake system with non-natural, structurally rather simple siderophores is a viable and efficient strategy to visualize infections in vivo.Compared to previous siderophore-PET probes, the separation of iron and PET-tracer binding sites allows to also accommodate other metal cations (e.g. 111 In or 64 Cu) that might fit less well into the siderophore binding site that was optimized for iron.The versatile scaffold, well-amenable to further modifications, 34,63 allows to fine-tune properties, and/or to introduce additional functionalities like an antibiotically active moiety to obtain full bacteria-targeted theranostics.
Indeed, a structural optimization is indicated to further improve parameters like tracer stability in vivo or enrichment in bacteria.The findings also pave the way for siderophore-based PETimaging in larger, non-rodent species.

EXPERIMENTAL SECTION Chemical Synthesis
Chemical and reagents were purchased from commercial vendors (TCI, Carl Roth, Baker and Sigma-Aldrich), if not stated otherwise, and employed without further purification in the below synthetic procedures.For synthesis, solvents with purity grade 99.5%, extra dry, absolute, AcroSeal TM , ACROS Organics TM were used.Work up procedures and purifications solvents were either HPLC or p. A. grade.Glassware was oven-dried prior to synthesis.Reaction progress was controlled by thin layer chromatography (TLC) or Liquid Chromatography-coupled Mass Spectrometry (LCMS).All compounds had purity ≥95% as determined by high-performance liquid chromatography (UV detection) and 1 H-/ 13 C-NMR analysis.

benzyl (2-(2-bromoacetamido)ethyl)carbamate (1)
According to a literature procedure from K. Ferreira To a white suspension of N-Cbzethylendiamine (504 mg, 2.6 mmol, 1.0 eq) in DCM (1 mL) a solution of K2CO3 (786 mg, 5.7 mmol, 2.2 eq) in MilliQ H2O (4 mL) and a solution of bromo acetyl bromide (270 µL, 3.1 mmol, 1.2 eq) in DCM (4 mL) was added dropwise at 0 °C.The solution was stirred vigorously at 0 °C.The reaction was equilibrated to 21 °C and continued stirring at 21 °C for 1h.The phases were separated, and the organic phase was dried over Na2SO4 and concentrated in vacuo.The crude product 1 was obtained as a white solid and used without any further purification (715 mg, acc. to LCMS min.96%, 87%).
The combined organic extracts were washed with brine (100 mL) and dried over Na2SO4, filtered and concentrated in vacuo.The obtained clear oil 3 (302 mg, 0.58 mmol, 98%) was used without any further purification.Compound 4 (145 mg, 194 µmol, 1.0 eq) was dissolved in anhydrous MeOH (0.5 mL), degassed with Argon balloons and Pd/C (6 mg, 57 µmol, 0.3 eq) was added under an Argon atmosphere.Subsequently, H2 filled balloons were inserted into the reaction solution and the reaction was stirred for 2 h at 23 °C.Then the catalyst was removed by filtration, the filtrate was concentrated to dryness via rotary evaporation and crude product 5 was obtained as a yellow solid (112 mg, 182 µmol, 92%), which was employed in the next step without further purification.To a solution of acetylated compound 7 (44 mg, 70 µmol, 1.0 eq) in anhydrous MeOH (0.8 mL), cooled to 0 °C, DIPEA (0.2 mL) was added, the 20% solution was equilibrated to 22 °C and continued stirring for 4 hours.After removal of the solvent and purification by RP-HPLC (C18-phenomenex, 40 min gradient: 0-20% MeCN/H2O 0.1% HCOOH) and lyophilization product 8 could be obtained as a white solid (22 mg, 38 µmol, 60%).
The phases were separated, and the aqueous phase was extracted with DCM (3 x 100 mL).

Figure 4 .
Figure 4. Growth recovery assay in siderophore-deficient E. coli mutant.(A) Principle of growth recovery assay, created using biorender and (B) growth recovery in the enterobactin (ENT) -deficient strain E. coli ΔentA in the presence of 10 µM compound ± 10 µM FeCl3 was assessed after incubation for 48 hours at 37 °C by OD600nm measurement in a plate reader.Bacteria were either grown in irondepleted (no iron) or 10 µM FeCl3-supplemented, phosphate-buffered LMR medium (n = 3).The growth relative to ENT is plotted; error bars correspond to ± standard deviation (SD).

Figure 6 :
Figure 6: Biodistribution of [ 68 Ga]7 (left side, A-C) and [ 68 Ga]15 (right side, D-E) in male C57Bl/6N mice over 60 min of dynamic PET imaging.(A+D) Representative time activity curves (TAC) display the biodistribution of [ 68 Ga]7 or [ 68 Ga]15 tracers for six organs (left and right kidney, bladder, liver, heart and muscle) in the course of one hour of dynamic PET/CT imaging (n = 6, error bars indicate ± SD.Detailed graphs showing all 32 frames can be found in S9 and S10.In (B+E) the % injected dose / organ weight during the last time frame (50-60 min) of dynamic PET/CT scan (%ID/g), is plotted, n=6, error bars indicate ± SD. (C+F) Tracer integrity and formation of hydrophilic metabolites in blood and urine samples (each n=6, error calculated as ± SD) after 60 minutes of dynamic PET/CT scan, determined by radio-HPLC measurements.

Figure 7 : 16 10
Figure 7: In vivo PET imaging in mice infected with E. coli with [ 68 Ga]7.(A) 3.0 x 10 7 CFUs E. coli (ATCC47076) were administered i.m. into the left leg, and 27 µg LPS were administered i.m. into the right leg 24 hours before imaging (n = 6 animals, male C57Bl/6N mice).[ 68 Ga]7 was prepared directly prior to the imaging experiment and injected i.v.into the tail vein.Dynamic PET scans were performed for 60 min with a using a micro PET/CT (transaxial view displayed, further data and video compilation are in the Supporting Information).(B) Time-activity curves (TAC) of decay-corrected [ 68 Ga]7, error bars are ± SD. (C) Dissected Musculus gastrocnemius after 60 min of PET/CT scan were imaged by autoradiography (AR), underwent Gram staining for bacteria (blue areas), immune histology with an anti-CD38 antibody (CD38) for macrophages (green fluorescence), and hematoxylin and eosin stain for cells and tissue.The upper row shows images of the E. coli-infected left muscle, while the lower row shows images of the LPS-injected right muscle.(D) The accumulated dose %ID/g of [ 68 Ga]7 over the course of the last 30 minutes of dynamic PET/CT scans was 1.9-fold higher in E. coli-infected compared to LPS-injected muscles (****: p-value <0.0001) 16

Figure 8 :
Figure 8: In vivo PET imaging in mice infected with E. coli with [ 68 Ga]15.(A) 2.2 x 10 7 CFUs E. coli were administered i.m. into the left leg, and 27 µg LPS was administered i.m. into the right leg 24 hours before imaging (n = 8 animals, male C57Bl/6N mice).[ 68 Ga]15 was prepared directly prior to the imaging experiment and injected i.v.into the tail vein.Dynamic PET scans were performed for 60 min with a using a micro PET/CT (transaxial view displayed, further data and video compilation are in the Supporting Information).(B) Time-activity curves (TAC) of decay-corrected [ 68 Ga]15, error bars are ± SD. (C) Dissected Musculus gastrocnemius after 60 min of PET/CT scan were imaged by autoradiography (AR), underwent Gram staining for bacteria (blue areas), immune histology with an anti-CD38 antibody (CD38) for macrophages (green fluorescence), and hematoxylin and eosin stain for cells and tissue.The upper row shows images of the E. coli-infected left muscle, while the lower row shows images of the LPS-injected right muscle.(D) The accumulated dose %ID/g of [ 68 Ga]15 over the course of the last 30 minutes of dynamic PET/CT scans was 1.7-fold higher in E. coli-infected compared to LPS-injected muscles (**: p-value <0.0035).