Arabinofuranose-derived PET radiotracers for detection of pathogenic microorganisms

Abstract

PURPOSE:

Detection of bacteria-specific metabolism via positron emission tomography (PET) is an emerging strategy to image human pathogens, with dramatic implications for clinical practice. In silico and in vitro screening tools have recently been applied to this problem, with several monosaccharides including L-arabinose showing rapid accumulation in E. coli and other organisms. Our goal for this study was to evaluate several synthetically viable arabinofuranose-derived ¹⁸F analogues for their incorporation into pathogenic bacteria.

PROCEDURES:

We synthesized four radiolabelled arabinofuranose-derived sugars: 2-deoxy-2-[¹⁸F]fluoro-arabinofuranoses (D-2-¹⁸F-AF and L-2-¹⁸F-AF) and 5-deoxy-5-[¹⁸F]fluoro-arabinofuranoses (D-5-¹⁸F-AF and L5-¹⁸F-AF). The arabinofuranoses were synthesized from ¹⁸F^- via triflated, peracetylated precursors analogous to the most common radiosynthesis of 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG). These radiotracers were screened for their uptake into E. coli and S. aureus. Subsequently the sensitivity of D-2-¹⁸F-AF and L-2-¹⁸F-AF to key human pathogens was investigated in vitro.

RESULTS:

All ¹⁸F radiotracer targets were synthesized in high radiochemical purity. In the screening study, D-2-¹⁸F-AF and L-2-¹⁸F-AF showed greater accumulation in E. coli than in S. aureus. When evaluated in a panel of pathologic microorganisms, both D-2-¹⁸F-AF and L-2-¹⁸F-AF demonstrated sensitivity to most gram-positive and gram-negative bacteria.

CONCLUSIONS:

Arabinofuranose-derived ¹⁸F PET radiotracers can be synthesized with high radiochemical purity. Our study showed absence of bacterial accumulation for 5-substitued analogues, a finding which may have mechanistic implications for related tracers. Both D-2-¹⁸F-AF and L-2-¹⁸F-AF showed sensitivity to most gram-negative and gram-positive organisms. Future in vivo studies will evaluate the diagnostic accuracy of these radiotracers in animal models of infection.

INTRODUCTION:

Bacterial infection is a major health threat in the United States and worldwide. According to the Center for Disease Control and Prevention (CDC), more than 2 million people get infected and approximately 23,000 die of bacterial infections each year in United States, with special risks posed by hospital-acquired and antibiotic-resistant microorganisms[1]. Accurate and rapid diagnosis of infection will reduce patient morbidity through effective therapeutic intervention. In addition to traditional laboratory and sampling methods, diagnostic imaging tools are frequently used in the work-up of infection and include computed tomography (CT), single photon emission CT (SPECT), magnetic resonance imaging (MRI) and hyperpolarized ¹³C magnetic resonance (HP MR) spectroscopy. While essential, these methods frequently suffer from low diagnostic accuracy due to their inability to differentiate infection from inflammation, cancer, and rheumatologic disease, and tedious data acquisition [2–5]. In light of these concerns, several groups including ours hope to develop faster, more accurate, and higher resolution positron emission tomography (PET) tools for imaging bacterial infection. Newer radiotracers derived from maltose, para-aminobenzoic acid (PABA), D-amino acids, bacterial siderophores, sorbitol, and other small-molecules, have targeted bacteria-specific metabolism, with the potential for widespread clinical dissemination [6–12]. The sorbitol-derived radiotracer 2-deoxy-2-[¹⁸F]fluoro-sorbitol ([¹⁸F]FDS), a reduced product of [¹⁸F]FDG, has been studied extensively in elegant preclinical models and more recently in patients[13–15]. This radiotracer shows high uptake in key strains of gram-negative bacteria including multidrug resistant E. coli, K. pneumonia, and Y. enterocolitica.

As shown by [¹⁸F]FDS, exploiting the metabolic requirements of pathogenic bacteria for various sugar alcohols can drive the discovery of new PET tracers targeting bacterial infection. In silico and in vitro screening methods for bacteria-specific sugars, employing commercially-available β-emitting ¹⁴C and ³H molecules, have identified numerous other candidate probes including arabinose-derived structures. Specifically, [1-¹⁴C]L-arabinose showed avid and selective accumulation in E. coli[16]. In gram-negative bacteria, both D- and L-arabinose are converted into ribulose by the action of arabinose isomerase (Figure 1). Next, ribulokinase phosphorylates the oxygen at the 5-position generating ribulose-5-phosphate. Finally, the action of ribulose-5-phosphate-4-epimerase results in xylulose-5-phosphate, which enters the pentose phosphate pathway (PPP)[17–19]. In several bacteria, additional metabolic pathways may be employed to incorporate D- and L-arabinose and their metabolites. For example, in gram-positive bacteria clostridium tetani, D-arabinose-5-phosphate (A5P) isomerase catalyzes D-ribulose-5-phosphate conversion into D-arabinose-5-phosphate[20]. Beyond transport, biotransformation, and phosphorylation, there are additional mechanisms of bacteria-specific probe retention for L and D-arabinose derived analogues. For example, the L-arabinose binding protein (ABP) derived from the E. coli periplasm can bind both α- and β- anomers of L-arabinose. The ABP acts as a primary receptor for L-arabinose accumulation by bacteria [21–24].

Figure 1. Potential mechanisms of arabinofuranose incorporation by bacteria.

(A) Structures of ¹⁸F D- and L-arabinofuranose analogs studied and (B) metabolism of D- and L-arabinofuranose highlighting the metabolic fates of the 1–5 positions of these molecules. Arabinose isomerases (E.C. 5.3.1.3 and 5.3.1.4 for D and L forms respectively) convert D-(and L-) arabinoses into D-(and L-) ribuloses, which are subsequently phosphorylated at 5-position hydroxyl group catalyzed by ribulose kinase (E.C. 2.7.1.47 and 2.7.1.16 for D and L forms respectively). Next, ribulose-5-phosphate epimerase (E.C. 5.1.3.1 and 5.1.3.4 for D and L forms respectively) yields 2-D-xylulose-5-phosphate, which finally enters the pentose phosphate pathway. Of note both 2- and 5-¹⁸F substitution yields ¹⁸F structures that are poor substrates for arabinose isomerase and ribulose kinase respectively; radiotracer incorporation for these analogs is likely mediated by transport and arabinose binding.

We hypothesized that different structural isomers of ¹⁸F-labeled arabinofuranose would incorporate differentially into bacteria. In the context of high bacterial accumulation, these radiotracers could potentially be used as PET probes to distinguish acute bacterial infection from radiologic mimics. The radiotracers described in this report could also potentially be applied to imaging fungi and mammalian cells that metabolize arabinofuranoses. In this work, we developed effective radiosyntheses of four arabinofuranose-derived PET radiotracers, namely 2-deoxy-2-[¹⁸F]fluoro-D-arabinofuranose (D-2-¹⁸F-AF), 2-deoxy-2-[¹⁸F]fluoro-L-arabinofuranose (L-2-¹⁸F-AF), 5-deoxy-5-[¹⁸F]fluoro-D-arabinofuranose (D-5-¹⁸F-AF), and 5-deoxy-5-[¹⁸F]fluoro-L-arabinofuranose (L-5-¹⁸F-AF). These radiotracers were subsequently screened for bacterial incorporation in vitro.

EXPERIMENTAL:

Syntheses of ¹⁹F arabinofuranose standards:

5-deoxy-5-fluoro-D-arabinofuranose and 5-deoxy-5-fluoro-L-arabinofuranose (D-5-F-AF and L-5-F-AF):

For detailed descriptions of syntheses for all new compounds, please refer to the Supporting Information. The compound D-5-F-AF was synthesized according to a previously published procedure [25]. The compound L-5-F-AF (5L-5, see B.1. in Supporting Information) was synthesized using an analogous method.

2-deoxy-2-fluoro-D-arabinofuranose and 2-deoxy-2-fluoro-L-arabinofuranose (D-2-F-AF and L-2-F-AF):

Compound D-2-F-AF was synthesized according to a previously published procedure [26]. (2L-3, see B.2. in Supporting Information).

Radiosyntheses of ¹⁸F arabinofuranoses:

D-5-¹⁸F-AF and L-5-¹⁸F-AF:

These radiotracers were synthesized in two steps (see C.2. in Supporting Information). First, the cyclotron derived ¹⁸F in [¹⁸O]H₂O was passed through a QMA anion exchange column eluted with 1 mL solvent (500 μL water + 500 μL acetonitrile) K₂CO₃ (2 mg) and kryptofix K222 (12 mg). This mixture was dried under nitrogen gas and vacuum with multiple additions of anhydrous acetonitrile (4X) (K₂CO₃- kryptofix K222 mix). The compound 5D-2’ or 5L-2’ (5 mg) was dissolved in anhydrous DMSO (500 μL) and added to the above anhydrous K₂CO₃- kryptofix K222 mix and heated to 155 °C for 10 min. The R_f value of the ¹⁸F labelled compounds by radioTLC (1:1 hexanes:ethyl acetate) were comparable with the R_f values of 5L-4 and its D-counterpart standards (see Figure S1A and S2A for radioTLC), and HPLC analysis of the ¹⁸F compounds with co-injected characterized standards confirmed the identity of the esterified D- and L-5-¹⁸F sugar alcohols (see Figure S1C, D, E and S2C, D, E for analytical HPLC). Second, the acetate groups were deprotected in a C-18 plus short column via 1N NaOH. Excess ¹⁸F ion was removed by passing the final compound through C-18 plus short column followed by 2 x(AG11-A8-long aluminaN) columns. RadioTLC (95:5 CH₃CN:H₂O) confirmed the synthesis of D-5-¹⁸F-AF and L-5-¹⁸F-AF (see Figure S1B and S2B for radioTLC). The R_f value of the final products was consistent with the R_f value of the cold standard (5L-5).

D-2-¹⁸F-AF and L-2-¹⁸F-AF:

These syntheses were performed similarly and are fully described in section C.4. of the Supporting Information. Briefly, anhydrous ¹⁸F-K₂CO₃-K222 mixture was refluxed with compound 2D-2’ (or 2L-2’) in dry acetonitrile (500 μL). Both analytical HPLC and RadioTLC (1:1 hexane:ethyl acetate) confirmed the ¹⁸F labelling (see Figure S3A and S4A for radioTLC) In both cases, the ¹⁸F labeled intermediates co-eluted with the standard D- (or L-) 2-F-OBz₃-α-AF (see Figure S3C, D, E and S4C, D, E for analytical HPLC). The benzoyl esters were subsequently removed on C-18 plus short column using 1N NaOH (1 mL) for 3 min. This crude reaction mixture was neutralized with 1N HCl and finally passed through a sequence of C-18-AG11-A8-long aluminaN-long aluminaN columns. The purity of the aliquots was determined in RadioTLC (95:5 CH₃CN:H₂O) (see Figure S3B and S4B for radioTLC).

In vitro analyses of ¹⁸F-arabinofuranoses:

Screening studies were performed to evaluate ¹⁸F arabinofuranose tracer accumulation by E. coli and S. aureus. E. coli and S. Aureus were grown aerobically in LB for 16 hours with agitation of 111 rpm. The cultures were pelleted at 3400 rpm for 5 minutes and resuspended in an equivalent amount of Ham’s F12 media (Gibco). Following a 1/16 dilution, cultures were then incubated with 1 μCi of ¹⁸F arabinofuranoses (D-2-¹⁸F-AF, L-2-¹⁸F-AF) at 180 rpm for 120 minutes. The bacterial suspensions were transferred to filter tubes (Corning Costar Spin-X) and centrifuged at 8000 rpm for 5 minutes. Phosphate buffed saline was added to each tube and the cultures were centrifuged at 8000 rpm for 5 minutes. The pellet and supernatant were separated and counted on a γ counter (Hidex Automatic Gamma Counter). Four replicates were performed for each bacterial strain.

A sensitivity study was subsequently performed to evaluate ¹⁸F arabinofuranose accumulation by a panel of disease-relevant pathogens. The bacterial strains used, and their growth conditions are listed in Table S1. Bacteria strains (except M. marinum) were grown aerobically in their listed medias for 16 hours with agitation of 111 rpm. M. marinum was grown aerobically for 3 days with media replenishment every 24 hours. Culture and other experimental methods were identical to those described above. Again, four replicates were performed for each bacterial strain.

Statistical analyses:

All synthetic data including radiochemical yields, and % radiochemical purities are reported as mean ± standard error. Radiochemical yields are reported with and without decay-correction for ¹⁸F (t_1/2=110min). In vitro data were normalized to OD₆₀₀ for sensitivity analysis to account for differential growth rates between organisms. All statistical analysis was performed using Microsoft excel and Prism 8.2. Four data sets were acquired for all in vitro studies (N= 4). Data were analyzed using an unpaired two-tailed Student’s t-test. All graphs are depicted with error bars corresponding to the standard error of the mean.

RESULTS AND DISCUSSION:

Radiosyntheses of D-2-¹⁸F-AF, L-2-¹⁸F-AF, D-5-¹⁸F-AF, L-5-¹⁸F-AF and their corresponding ¹⁹F standards.

To investigate bacterial incorporation in vitro, four ¹⁸F-labeled arabinofuranose sugars were elaborated. Detailed synthetic and radiosynthetic procedures, as well as compound characterization are described in the Supplemental Information. The basic strategy involved synthesis of triflated, peracetylated precursors for each molecule, to mimic the most common synthetic method used for [¹⁸F]FDG. For D-5-¹⁸F-AF and L-5-¹⁸F-AF, the increased nucleophilicity of the hydroxyl at the 5-position was used to generate orthogonally protected precursors (Figure 2A). A previous synthesis of D-2-¹⁸F-AF was adapted for this purpose, while an analogous method was used for its enantiomer L-2-¹⁸F-AF (Figure 2B). Diethylaminosulfur trifluoride (DAST) chemistry was then used to generate all ¹⁹F standards which were characterized fully using ¹H, ¹³C, ¹⁹F NMR and high-resolution mass spectrometry. Subsequent radiosynthesis using nucleophilic ¹⁸F was used to synthesize the four sugars, which were isolated via cartridge purification and characterized via radio-TLC (Figure S1A, B; S2A, B; S3A, B and S4A, B). The summary of characterization data is as follows: D-2-¹⁸F-AF (N=12), decay-corrected yield (13.2 ± 1.6)%, end of synthesis (EOS) radiochemical yield (7.5 ± 1.2)%, radiochemical purity (98.0 ± 2.0)%; L-2-¹⁸F-AF (N=10, decay-corrected yield (18.5 ± 2.6)%, EOS radiochemical yield (10.5 ± 1.4)%, radiochemical purity (98.0 ± 2.0)%; D-5-¹⁸F-AF (N=3, decay-corrected yield (2.0 ± 0.2)%, EOS radiochemical yield (1.0 ± 0.1)%, radiochemical purity (93.0 ± 1.8)%; L-5-¹⁸F-AF (N=3, decay-corrected yield (2.2 ± 0.2)%, EOS radiochemical yield (1.1 ± 0.1)%, radiochemical purity (94.8 ± 1.3)%.

Figure 2. Synthesis of ¹⁸F-labelled arabinofuranose-derived structures.

A) D-5-¹⁸F-AF and L-5-¹⁸F-AF radiosyntheses B) D-2-¹⁸F-AF and L-2-¹⁸F-AF radiosyntheses

Screening of ¹⁸F arabinofuranoses revealed accumulation of D-2-¹⁸F-AF and L-2-¹⁸F-AF in E. coli.

Two of the most important bacterial pathogens are E. coli and S. aureus, which cause several clinically-relevant infections. Specifically, gram-negative E. coli is a frequent cause of urinary tract, intestinal, and biliary infections [27, 28]. S. aureus is a frequent gram-positive organism cultured in skin, musculoskeletal, and blood-borne infections [29, 30]. Therefore we initially evaluated ¹⁸F arabinofuranoses in these organisms prior to a broader survey of pathogens (Figure 3). Initial studies showed low accumulation of all radiotracers into S. aureus (mean uptake < 2.0 Bq/10 million cells). Furthermore, there was lower accumulation of L-5-¹⁸F-AF and D-5-¹⁸F-AF in both E. coli and S. aureus (mean uptake < 2.0 Bq/10 million cells). Higher incorporation was observed for L-2-¹⁸F-AF and D-2-¹⁸F-AF in E. coli; similar uptake was found for both radiotracers in E. coli, significantly higher than all probe combinations (P < 0.05 in all cases). The accumulation of L-2-¹⁸F-AF in E. coli was 6-fold higher (P < 0.0001) than that in S. aureus, and 2.5-fold higher than that of its corresponding 5-labeled counterpart (P = 0.0001). For D-2-¹⁸F-AF, accumulation in E. coli was 4.5-fold higher than that in S. aureus, and 2.4-fold higher than that of its corresponding 5-labeled counterpart (P = 0.0072). The specificity of this process for L-2-¹⁸F-AF was further validated in blocking studies (Figure 4A,B). In summary, these studies indicated higher accumulation in gram-negative organisms for D-2-¹⁸F-AF and L-2-¹⁸F-AF and motivated further sensitivity studies.

Figure 3. Screening of arabinofuranose-derived ¹⁸F PET radiotracers in E. coli and S. aureus.

In vitro uptake of L-2-¹⁸F-AF in live E. coli was approximately 6-fold higher than that in S. aureus (P <0.0001). D-2-¹⁸F-AF accumulates in live E. coli approximately 4-fold more than in live S. aureus (P=0.0002). In vitro uptake of D-5-¹⁸F-AF and L-5-¹⁸F-AF showed <1.5 Bq/10 million bacterial cells incorporation for live E. coli and S. aureus in all cases. There was no statistically significant difference in accumulation, for any combination of D- and L-5-¹⁸F-AF radiotracer and organism.

Figure 4. Blocking studies for L-2-¹⁸F-AF.

A) L-2-¹⁸F-AF accumulates in live E. coli 2.5x more than in heat-killed (P <0.0001), and 1.75x more than in a blocking study using 10 mM unlabeled L-arabinose (P=0.0006). This probe is retained in live E. coli approximately 4 times more than in live S. aureus. B) L-2-¹⁸F-arabinofuranose uptake in E. coli in competition with unlabeled L-arabinofuranose at various concentrations. Cellular uptake of the radiotracer decreases as the concentration of L-arabinofuranose increases in the solution.

The sensitivities of D-2-¹⁸F-AF and L-2-¹⁸F-AF were evaluated in vitro using a broad panel of important human pathogens.

Based on screening data, D-2-¹⁸F-AF and L-2-¹⁸F-AF were studied in a panel of pathogenic bacteria. These bacteria were chosen as representative gram-negative and gram-positive organisms implicated in a variety of dangerous human infections. For example, gram-negatives K. pneumoniae and P. aeruginosa are involved in hospital-acquired disease while gram-positive S. epidermidis is a cause of skin and catheter-related infections [31, 32]. A reproducible, high-throughput screening assay was developed for simultaneous evaluation of 12 living microorganisms using a single ¹⁸F radiotracer synthesis. Both D-2-¹⁸F-AF and L-2-¹⁸F-AF were incorporated into several gram-negative and gram-positive bacteria as shown in Figure 5A,B. For D-2-¹⁸F-AF the highest radioactivity was retained for P. mirabilis, S. typhimurium, E. faecalis and S. epidermidis (approximately 7–8 Bq/10 million cells). The uptake levels of D-2-¹⁸F-AF in these four bacteria were similar (< 3 fold difference based on incorporated radioactivity/10 million cells) to data recently obtained for D-[methyl-¹¹C]methionine [33]. Evaluation of L-2-¹⁸F-AF showed the highest uptake in K. pneumoniae, A. baumannii, and M. marinum which retained between 3.5 and 4 Bq/10 million cells. Of note D-2-¹⁸F-AF and L-2-¹⁸F-AF were incorporated into both gram-negative and gram-positive species. Also significantly, there was little or no signal retained for many important pathogens for example S. aureus, P. aeruginosa and L. monocytogenes.

Figure 5. Sensitivity study in a panel of bacterial pathogens.

A) D-2-¹⁸F-AF showed the highest retention in P. mirabilis, S. typhimurium and E. faecalis. B) L-2-¹⁸F-AF showed the highest retention in several gram-positive and gram-negative bacterial strains with K. pneumoniae, A. baumannii and M. marinum accumulating about 4 Becquerel/10 million cells.

CONCLUSIONS:

Molecular imaging methods targeting bacteria-specific metabolism have outstanding clinical potential. This potential lies in distinguishing active infection from other disease entities that appear similar on morphologic imaging (CT and MRI), allowing appropriate antimicrobial management. With the advent of multi-modality scanning, PET is particularly well suited to complement the structural information typically obtained in the work-up of infected patients. In developing PET tracers for infection, both specificity and sensitivity are important. A PET tracer specifically targeting a microbial pathway is the simplest approach; however, another strategy might be based on relative avidity of the tracer for bacterial metabolism. For this reason, we considered both D- and L-arabinofuranose-derived PET tracers.

Both D- and L-arabinofuranose derived ¹⁸F radiotracers were obtained in high radiochemical purity. Interestingly, our screening assay showed little or no bacterial incorporation of ¹⁸F-arabinofuranoses with substitution at the 5-position. Although mechanistic analysis is beyond the scope of this manuscript, this finding suggests the 5-OH is necessary for retention in bacteria, perhaps based on interaction with arabinose-binding proteins. In contrast the 2-substituted molecules D-2-¹⁸F-AF and L-2-¹⁸F-AF demonstrated incorporation into live E. coli, which was not seen in heat-killed bacteria or in the presence of the non-labeled parent sugars. The mechanisms of retention for these two tracers are currently unknown and the basis for future study. As highlighted by Figure 1, ¹⁸F substitution at the 2-position would render D-2-¹⁸F-AF and L-2-¹⁸F-AF poor substrates for arabinose isomerase (E.C. 5.3.1.3 and 5.3.1.4 for D and L forms respectively). In this case the bacterial retention of these probes may be mediated by their relative affinities for L-arabinose binding proteins. Additional arabinofuranose-derived ¹⁸F substrates that could potentially undergo biotransformation would be 3-deoxy-3-[¹⁸F]fluoro-L-arabinofuranose and 3-deoxy-3-[¹⁸F]fluoro-D-arabinofuranose. We made several attempts to synthesize these using an identical radiochemical strategy, without success. Several factors can impede a ring S_N2 fluorination reaction, including steric hindrance from neighbouring positions, ring conformations that impede nucleophilic attack, and neighbouring hydroxyl protecting groups that favor elimination [34, 35]. These 3-position arabinofuranose-derived analogues are a topic of future study.

A bacteria-targeted PET tracer would be especially valuable clinically in two scenarios, namely (1) if the tracer detects all or most pathogenic microorganisms, allowing identification of infection versus other processes (2) if the tracer detects important categories of organisms that are relevant to antimicrobial therapy (e.g. [¹⁸F]FDS which is sensitive to most Enterobacteriaceae). The radiotracers investigated in this manuscript D-2-¹⁸F-AF and L-2-¹⁸F-AF showed promising sensitivity to several microorganisms, with no apparent preference for gram-negative vs. gram-positive bacteria. Furthermore, there were several important pathogens that showed low accumulation of these radiotracers most notably S. aureus. Future studies will investigate whether the species-specific incorporation of D-2-¹⁸F-AF and L-2-¹⁸F-AF can be used to non-invasively identify strains, and represents an advantage in compelling clinical scenarios. Importantly, these radiotracers may also have significant application to other organisms and diseases; arabinofuranoses are metabolized avidly by fungi [36, 37], and as highlighted by previous work the D-arabinose isoforms in particular may find use in oncologic imaging.