Pathogen-specific Bacterial Imaging in Nuclear Medicine

Abstract

When serious infections are suspected, patients are often treated empirically with broad-spectrum antibiotics while awaiting results that provide information on the bacterial class and species causing the infection, as well as drug susceptibilities. For deep-seated infections, these traditional diagnostic techniques often rely on tissue biopsies to obtain clinical samples which can be expensive, dangerous and has the potential of sampling bias. Moreover, these procedures and results can take several days and may not always provide reliable information. This combination of time and effort required for proper antibiotic selection has become a barrier leading to indiscriminate broad-spectrum antibiotic use. Exposure to nosocomial infections and indiscriminate use of broad-spectrum antibiotics are responsible for promoting bacterial drug-resistance leading to substantial morbidity and mortality, especially in hospitalized and immunosuppressed patients. Therefore, early diagnosis of infection and targeted antibiotic treatments are urgently needed to reduce morbidity and mortality caused by bacterial infections worldwide. Reliable pathogen-specific bacterial imaging techniques have the potential to provide early diagnosis and guide antibiotic treatments.

INTRODUCTION

Bacterial infections are a major cause of morbidity and mortality worldwide. However, with antimicrobial resistance emerging and spreading globally, multi-drug resistant organisms (MDRO) are an increasingly serious threat to human health. In the United States alone, at least 2 million people become infected with MDRO bacteria each year, with at least 23,000 of these patients dying as a direct result of these infections, leading to significant social and economic costs.^1–3 Although antimicrobial resistance can occur naturally, the misuse and overuse of antibiotics is accelerating this process.^4–8 Improved diagnostic accuracy of pathogen-specific imaging techniques could curb the inappropriate use of antibiotics for non-infectious entities, streamline empiric antibiotic selection to target the class of the bacteria causing the infection and provide a way to rapidly monitor antibiotic treatments.

A common clinical problem for practicing infectious disease specialists and radiologists is differentiating active infection from other causes of inflammation. Current practice in microbiology (microbiologic cultures or molecular techniques) relies on the availability of a clinical sample (blood, urine, etc.). However, when a deep-seated infection is suspected, a biopsy is often required, leading to costly, invasive and sometimes dangerous procedures, prone to sampling errors. Therefore, clinicians frequently use conventional radiological techniques such as computerized tomography (CT), magnetic resonance imaging (MRI), X-ray, and ultrasound, to aid in the diagnosis and localization of deep-seated infections.⁹ However, these techniques rely on the presence of structural abnormalities in tissues and organs caused by the infection and/or inflammatory response of the host and are unable to reliably differentiate infection from inflammation or oncologic processes. To try to address this clinical challenge, multiple molecular imaging techniques have been developed to detect bacteria in vitro, in animal models, and for clinical studies. Extensive research has been done using optical imaging techniques with promising results,^10–14 and although these studies have helped significantly in understanding disease pathogenesis and in developing therapeutics for infections, their clinical evaluation for bacterial infections has been limited by tissue penetration. Conversely, nuclear imaging techniques such as positron emission tomography (PET) and single-photon emission computed tomography (SPECT) are not limited by tissue depth and use very small amounts of radioactive tracers that reflect the physiological and biochemical changes associated with the infection process.

Nuclear imaging of bacterial infections has been a developing field for more than 50 years and provides valuable insight not only in the diagnosis and monitoring of infections, but it can also help understand the disease pathogenesis in both preclinical and clinical settings.¹⁵ Although most of the published literature in this area involves non-specific imaging agents such as ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG), ⁶⁷Ga-citrate and in vitro radiolabeled leukocytes, significant efforts have been made to develop bacteria-specific radiolabeled tracers that can bind or accumulate only in bacteria.¹⁶ This is a challenging task since the ideal bacteria-specific imaging compound should be both sensitive and specific for detection of the infection site, in addition to being nontoxic, affordable, widely available, and easily and rapidly prepared (Table 1). These compounds should have ideal biochemical properties (e.g. moderate lipophilicity, low plasma protein binding, and metabolic stability),^17,18 as well as optimal pharmacokinetic and pharmacodynamic features for maximum performance and usability.¹⁹ The tracer should be able to rapidly localize to the infection site with little non-specific binding and have an optimal clearance rate that allows sufficient time for it to reach its target but is rapid enough to excrete unbound probe and minimize radiation exposure.²⁰ The search for the best bacteria-specific imaging tracers is still an ongoing challenge and this review summarizes the efforts made to achieve this goal.

Table 1.

Desirable properties of an ideal bacteria-specific PET imaging tracer (adapted from Gemmel et al.¹⁹ and Johnson et al.¹⁸)

Sensitive	High target to background signal ratio. Low limit of detection.
Specific	Probe targets infectious, not inflammatory lesions.
Quantitative	Signal proportionate to infectious burden.
Rapid	Fast localization to the site of infection.
Stable	Probe retained at the site of infection. No degradation of the probe by the host.
Safe	Acceptable radiation dose. Repeat injection feasible, without pharmacologic or immunologic effect.
Manufacturable	Uncomplicated synthesis with reasonable expense.

METABOLISM-BASED TRACERS

Based on the principles utilized in clinical microbiology, which uses selective metabolism to differentiate microbes,²¹ bacteria-specific imaging tracers based on selective metabolic pathways expressed only in bacteria (or a specific class of bacteria) could be developed not only to detect the presence of bacterial infection, but also to identify the class of bacteria causing the infections.²² Metabolism-based tracers, such as radio-fluorinated sugars and sugar alcohols, combined with favorable radiation dosimetry due to short half-life isotopes (e.g. ¹⁸F), provide multiple advantages such as easy penetration into diseased tissues, possible accumulation inside bacteria and rapid clearance from non-target tissues. However, since these tracers are designed to only be incorporated into bacteria, there would be limitations on their use for diagnosis of pathogens with a predominantly intracellular lifecycle (e.g. Listeria monocytogenes). Similarly, the incorporation rate of these tracers into the cells could be affected by the metabolic state of the bacteria.

Nucleoside analogs

The thymidine analog FIAU (1-(2′-deoxy-2′-fluoro-β-D-arabinofuranosyl)-5-iodouracil) was initially developed as a non-invasive agent to non-invasively monitor tumor-targeting bacteria. FIAU is a substrate for the native thymidine kinases (TK) of herpes viruses and a wide variety of bacterial species. Once inside the bacterial cell, FIAU gets phosphorylated by TK and subsequently trapped. However, bacteria that do not possess the TK enzyme (e.g. Pseudomonas aeruginosa), would not retain FIAU.²³ In 2004, Bettegowda et al. reported successful in vivo imaging of bacterial infections in mice using ¹²⁵I-FIAU.²⁴ Using a lung infection murine model, Pullambhatla et al. also showed that this tracer can be used to monitor the efficacy of antimicrobial therapy.²⁵ Subsequent studies with ¹²⁴I-FIAU reported promising results in patients with musculoskeletal infections (Fig. 1).²⁶ However, a recent study found that ¹²⁴I-FIAU lacks specificity in patients with prosthetic joint infections, with high background signal in uninfected muscle presumably due to host mitochondrial metabolism.²⁷ Although the synthesis methods and dosimetry of ¹⁸F-FIAU in animal models have also been reported, data for bacterial imaging is not available.²⁸

Figure 1. ¹²⁴I-FIAU signal in established infections as imaged by PET/CT.

Fused PET and CT images, taken at 2 hours after radiotracer administration, are shown for the following cases: (a) septic arthritis (right knee), (b) septic arthritis (right knee), (c) osteomyelitis (left distal tibia), (d) cellulitis (left lower extremity), (e) necrotizing septic arthritis (left knee). Reprinted with permission by the Public Library of Science from Diaz et al. 2007.²⁶

Sugars

While widely employed in clinical oncology, ¹⁸F-FDG is presently the most used PET tracer for imaging bacterial infections by identifying infection-associated inflammatory responses.^29,30 Immune cells increase the use of glucose as an energy source during metabolic bursts associated with inflammatory responses due to infection.³¹ Even during chronic inflammation, increased cell glycolysis persists,³² giving ¹⁸F-FDG the potential to image infections with high sensitivity.³¹ Current guidelines recommend ¹⁸F-FDG PET imaging in clinical cases of peripheral bone osteomyelitis, suspected spinal infection and fever of unknown origin (FUO),³³ and suggest it can be used in other infection settings. ¹⁸F-FDG PET has also been used to monitor tuberculosis (TB) patients,^34,35 evaluate patients with inconclusive diagnosis of infective endocarditis,³⁶ help distinguish between septic and aseptic loosening in joint prosthesis infection,³⁷ and diagnose diabetic foot infections.³⁸ However, since ¹⁸F-FDG is metabolized by the same pathways as glucose, and given the widespread expression of glucose surface transporters in eukaryotic cells, ¹⁸F-FDG is not able to differentiate true infection from sterile inflammation. Mills et al. developed the phosphate analog ¹⁸F-FDG-6-P in an effort to facilitate specific uptake of ¹⁸F-FDG by bacterial hexose phosphate transporters.³⁹ However, in vivo PET imaging with ¹⁸F-FDG-6-P in a foreign body Staphylococcus aureus infection mouse model demonstrated that the biodistribution was similar to that of ¹⁸F-FDG, therefore limiting its use for bacteria-specific imaging.

Sorbitol, a sugar alcohol, is a metabolic substrate for Enterobacteriaceae (Klebsiella pneumoniae, Escherichia coli, Yersinia spp., Enterobacter spp., etc.) which are the largest group of Gram-negative bacterial pathogens in humans. Similar to glucose, sorbitol is selectively taken up by bacteria via surface transporters, phosphorylated and further metabolized.⁴⁰ Mammalian cells do not have transporters for this sugar.⁴¹ The first approach to radiolabel a sorbitol analog was done by Li et al. who used a 30-min reaction to reduce ¹⁸F-FDG into ¹⁸F-fluorodeoxysorbitol (¹⁸F-FDS).⁴² Subsequently, Weinstein and Ordonez et al. demonstrated that ¹⁸F-FDS rapidly and selectively accumulated in Enterobacteriaceae (~1000 fold higher than mammalian cells), including MDR clinical strains, but not in Gram-positive bacteria or mammalian/cancer cells (Fig. 2).⁴³ In a murine myositis model, ¹⁸F-FDS PET rapidly differentiated infection from sterile inflammation with a limit of detection of 6.2 ± 0.2 log₁₀ colony forming units (CFU) for E. coli. This is a promising threshold since clinically relevant infections have much higher burdens of approximately 8 log₁₀ CFU bacteria per mL,⁴⁴ with volumes of tens to hundreds of milliliters.^45,46 These findings have been extended to models of mixed Gram-positive and Gram-negative thigh co-infections, brain infection, mice undergoing immunosuppressive chemotherapy, and K. pneumoniae pneumonia.^43,47,48 Importantly, ¹⁸F-FDS PET also monitored the efficacy of antimicrobial treatment with the PET signal correlating with the bacterial burden. Recently, ¹⁸F-FDS was determined to be safe and well tolerated after a single intravenous dose injected into healthy human volunteers to assess biodistribution and radiation dosimetry.⁴⁹

Figure 2. ¹⁸F-FDS PET/CT imaging of E. coli myositis in immunocompetent mice.

¹⁸F-FDS signal is noted in the infected (yellow arrow) but not in the inflamed (control) sterile thigh (red arrow). ¹⁸F-FDG signal was noted in both infected and inflamed thighs. Reprinted with permission by the American Association for the Advancement of Science from Weinstein and Ordonez et al. 2014.⁴³

Maltose and maltodextrin are polysaccharides consisting of two or multiple glucose units respectively. These sugars are incorporated with high specificity using the maltose-maltodextrin transport system, present in multiple Gram-negative and Gram-positive bacterial species, but not in mammalian cells.⁵⁰ In 2011, Ning et al. reported maltodextrin-based optical imaging probes that could specifically-detect E. coli myositis in mice with high specificity.⁵¹ Subsequent work led to the development of ¹⁸F-labeled maltohexaose (MH¹⁸F).⁵² Using ¹⁹F NMR spectroscopy, MH¹⁹F was found to have high specificity for bacteria over mammalian cells, accumulating two times more in E. coli compared to hepatocytes. Using a rat E. coli myositis model, MH¹⁸F PET was able to localize the infected tissue as early as 10 min post-injection, and 60 min later the signal in infected muscle was 8.5-fold higher compared to uninfected control muscle. Unlike ¹⁸F-FDG, MH¹⁸F was able to differentiate between live (infection) and dead (inflammation) bacteria in muscle. MH¹⁸F was also used to effectively measure drug resistance and monitor the therapeutic effect of antibiotics in vivo. Therefore, MH¹⁸F could be a useful PET imaging agent for a broad spectrum of Gram-positive and Gram-negative infections. A similar approach was adopted by Gowrishankar et al. in the development of 6-¹⁸F-fluoromaltose, which accumulated in in vitro cultures of E. coli, P. aeruginosa and L. monocytogenes but not in mammalian cells.⁵³ PET imaging with this compound was also able to differentiate infected from inflamed muscle. However, 6-¹⁸F-fluoromaltose had suboptimal pharmacokinetics and poor signal-to-background ratios, leading to the development of the second-generation tracer 6″-¹⁸F-fluoromaltotriose, with promising results detecting E. coli myositis and P. aeruginosa wound infections in vivo (Fig. 3).⁵⁴

Figure 3. 6″-¹⁸F-fluoromaltotriose PET/CT imaging of E. coli myositis.

Three-dimensional color map from small-animal PET/CT scan of representative nude mouse with E. coli infection in its right thigh 24 h after infection and 1 h after intravenous injection of 7.4 MBq of 6″-¹⁸F-fluoromaltotriose. Also seen in image is bladder (BL) and kidneys (K). Yellow arrow highlights site of infection. This research was originally published in JNM. Gowrishankar, G. et al. J Nucl Med. 2017;58:1679–1684. © by the Society of Nuclear Medicine and Molecular Imaging, Inc.⁵⁴

Trehalose is a disaccharide essential for mycobacterial growth and virulence⁵⁵, and there is evidence of the uptake of trehalose analogs by Mycobacterium tuberculosis.⁵⁶ Recently, Rundell et al. reported that a fluorinated trehalose analog (¹⁹F-FDTre) accumulates in Mycobacterium smegmatis and provided a synthetic scheme for a ¹⁸F-radiolabeled version.⁵⁷ Although these results are still preliminary, they suggest the potential of radiolabeled trehalose analogs for Mycobacteria-specific imaging.

Cell wall components

Amino sugars are essential components of peptidoglycan, the strong polymer of the bacterial cell wall that maintains the cell shape and anchors components of the cell envelope. The peptidoglycan is built from alternating units of N-acetylglucosamine and N-acetylmuramic acid, cross-linked with oligopeptides at the lactic acid residue of N-acetylmuramic acid.⁵⁸ Using a N-acetylglucosamine derivative, Martinez et al. reported the synthesis, preliminary characterization, and PET imaging of 2-deoxy-2-¹⁸F-fluoroacetamido-D-glucopyranose (¹⁸F-FAG) in E. coli infected rats.⁵⁹ The signal of ¹⁸F-FAG PET in infected muscle was 1.68 times higher compared to turpentine oil-induced inflammation. Another component of the peptidoglycan layer of the bacterial cell wall in both Gram-positive and Gram-negative bacteria are D-amino acids.^60,61 Unlike mammalian cells, most bacteria incorporate and produce significant amounts of D-amino acids into their cell walls,⁶² being the main target of a large number of antibiotics that antagonize the synthesis, dimerization and/or incorporation of D-amino acids. Using this approach, Neumann et al. recently reported the radio-synthesis of D-methyl-¹¹C-Methionine (¹¹C-D-Met), demonstrating rapid and selective differentiation of both E. coli and S. aureus from sterile inflammation in a myositis mouse model (Fig. 4).⁶³ Conversely, L-methyl-¹¹C-Methionine is not specific for infection imaging.^63,64 Preliminary studies have considered the potential of bacterial wall-labeling agents such as D-Alanine and D-Glutamate. However, significant in vivo defluorination for ¹⁸F-labeled versions of alanine,^65,66 modest enantiomeric excess for the reported ¹¹C-L-Alanine asymmetric synthesis,⁶⁷ and the lack of homology between reported ¹⁸F versions of glutamate and the native amino acid are significant drawbacks.^68,69 Previous studies have also indicated that many unnatural D-amino acids are incorporated into bacterial peptidoglycan with a high tolerance for structural differences, suggesting the existence of potential bacteria-specific imaging compounds that are yet to be explored.^70,71

Figure 4. In vivo studies using ¹¹C-D-Met in a murine myositis model.

Representative images show marked uptake in areas corresponding to live bacterial injection (left deltoid), in contrast to sterile inflammation (right deltoid) and normal muscle. The site of live bacterial inoculation is denoted by a red arrow, while the site of 10X heat-killed bacterial inoculation is denoted by a white arrow. Reprinted with permission by Springer Nature from Neumann et al. 2017.⁶³

Iron metabolism

For more than 40 years, ⁶⁷Ga-citrate SPECT has been extensively used to detect infection/inflammation.^29,72 Multiple pathways have been proposed for ⁶⁷Ga-citrate uptake in inflamed/infected tissues such as increased vascular permeability in the infection site,⁷³ direct uptake by bacteria,⁷⁴ uptake by leukocytes,⁷⁵ and binding to plasma transferrin, lactoferrin and/or bacterial siderophores.^76,77 Despite not being specific for bacterial infection, ⁶⁷Ga-citrate is still widely used around the world to identify sites of infection in patients with FUO.⁷⁸ Due to its shorter half-life, ⁶⁸Ga-citrate was developed to improve the long waiting time required for imaging after a ⁶⁷Ga-citrate injection.^{76 68}Ga-citrate has been evaluated in animal models of osteomyelitis and myositis ^79,80, as well as in patients with osteomyelitis.⁸¹ In patients with tuberculosis, ⁶⁸Ga-citrate accumulated in both pulmonary and extra-pulmonary tuberculous lesions.⁸² Based on the same principle of iron accumulation in infected tissues, Kumar et al. developed a ⁶⁸Ga-radiolabeled apo-transferrin complex (⁶⁸Ga-TF), which provided 7.5 times higher signal in the PET images of S. aureus muscle infection in rats, compared to uninfected uninflamed muscle.⁸³ No signal from the infected sites was evident within the first 2-hour post-injection in additional unconjugated ⁶⁸Ga-Cl₃ PET scans, suggesting some specificity of the transferrin-bound compound. Although these mechanisms are not bacteria-specific, there has also been interest in the use of pathogen-specific radiolabeled siderophores as potential bacteria-specific imaging agents.⁸⁴

ANTIMICROBIAL PEPTIDES

Peptides are short chains of amino acids linked by amide bonds. Antimicrobial peptides are synthesized by cells involved in the host immune response, and due to their cationic properties, they tend to bind more to the negatively charged bacterial cell wall compared to eukaryotic cells.^85,86 Based on the initial experiments with ubiquicidin (UBI), a human antimicrobial peptide present in the respiratory epithelium, Welling et al. synthesized and evaluated several ^99mTc-radiolabeled fragments of UBI with increased sensitivity and selectivity for bacteria.⁸⁷ Subsequent studies with ^99mTc-UBI (fragments 29–41) showed promising results imaging Gram-positive and Gram-negative bacteria, including MDR bacterial strains, and monitoring antimicrobial efficacy in both preclinical and clinical settings.^88–93 Based on the same peptide, Welling et al. developed ¹¹¹In-DTPA-Cy5-UBI(29–41), a hybrid fluorophore and radionuclide labeled antimicrobial peptide, which had a 2.82 and 2.37 target to background ratio in a S. aureus and K. pneumoniae myositis mouse model, respectively.^{94 18}F-UBI(29–41) has also been evaluated in a rat S. aureus myositis model with poor results due to extensive defluorination.⁹⁵ In 2014, Ebenhan et al. synthesized ⁶⁸Ga-NOTA-UBI(29–41) and showed specificity towards S. aureus in a rabbit model, being able to distinguish infection from sterile inflammation.⁹⁶ These findings were subsequently validated in a S. aureus mouse myositis model.⁹⁷ The first-in-human study evaluating the diagnostic performance of ⁶⁸Ga-NOTA-UBI(29–41) was performed in three patients with known/suspected bone-or soft tissue infection.⁹⁸ With a target to background ratio of 3.4 ± 0.2 at 60 minutes, ⁶⁸Ga-NOTA-UBI(29–41) was able to accurately identify the infection sites in these patients (Fig. 5). No adverse effects were reported with the administration of ⁶⁸Ga-NOTA-UBI(29–41).

Figure 5. ⁶⁸Ga-NOTA-UBI(29–41) image in a patient with peripheral bone and soft-tissue infection.

(A) Whole-body MIP-PET image shows a diffusely increased tracer uptake in the right lower leg. (B) Detailed PET-CT images demonstrate focal increased tracer uptake in the ankle-joint extending in the adjacent bone (black and white arrows) as well as diffuse tracer uptake in the calf muscles (red arrow) whereas there is no significant uptake in the contralateral leg. Images are obtained 60 min after tracer administration (240 MBq). This research was originally published in JNM. Ebenhan, T. et al. J Nucl Med. 2017. © by the Society of Nuclear Medicine and Molecular Imaging, Inc.⁹⁸

Using a derivative of depsidomycin, a peptide with significant antimicrobial activity, Mokaleng et al. developed and evaluated ⁶⁸Ga-DOTA-TBIA101 in a mouse model of E. coli myositis, where the tracer was able to differentiate infection from healthy tissue.⁹⁹ However, this same peptide accumulated in rabbit muscles infected with M. tuberculosis and sterile inflammation controls at the same rate, suggesting a non-specific uptake by the inflammation associated with the infection rather than by the bacteria.¹⁰⁰ In 2017, Dutta et al. radiolabeled and reported LL37, a human cathelicidin antimicrobial peptide (⁶⁸Ga-CDP1) that had higher uptake in in vitro bacterial cultures of S. aureus, E. coli and M. smegmatis, compared to eukaryotic cells.¹⁰¹ Additional not bacteria-specific radiolabeled peptides such as the human neutrophil peptide (^99mTc-HNP-1), neutrophil elastase inhibitor peptide (^99mTc-HNE-2), human β-defensin (^99mTc-HBD-3), and vascular adhesion protein-1 selective peptide (⁶⁸Ga-DOTAVAP-P1), have been evaluated for bacterial imaging with variable results.^102–105

ANTIBIOTICS

Antibiotics are compounds designed to kill (bactericidal) or impair the growth (bacteriostatic) of bacterial pathogens. Ideally, a radiolabeled antibiotic will bind to their bacterial targets with high specificity. However, an excess of unlabeled antibiotic may compete with the radiolabeled antibiotic or kill the pathogen, both of which will impair the imaging efficacy of the tracer. Additionally, desirable pharmacologic characteristics of antibiotics may not match those for radiolabeled bacterial imaging compounds (e.g. long half-life for convenient daily dosing). Moreover, bacterial resistance mechanisms to a specific antibiotic, such as efflux pumps, cell permeability changes, decreased target affinity or degradative metabolism can affect the radiolabeled analog and could limit its use as a diagnostic tracer. Radiolabeled antibiotics and PET imaging also provide a noninvasive technique for studying pharmacokinetics and pharmacodynamics in vivo.^106–110 A large number of antibiotics have been radiolabeled, with ^99mTc being the predominant radionuclide since the tracer preparation is easily accomplished using a cold kit format within 30 minutes.^111–113 However, a large number of these radiolabeled antibiotics have not been able to differentiate infection from inflammation and have not been evaluated in a clinical setting. While we provide a brief discussion of the most significant radiolabeled antibiotics, multiple extensive reviews are available on this topic.^{106,114–117}

Fluoroquinolones

Ciprofloxacin, a broad-spectrum bactericidal fluoroquinolone, was initially labeled with ^99mTc and evaluated in preclinical infection models with promising results.^{118 99m}Tc-Ciprofloxacin imaging was used in large clinical trials that reported high sensitivity and specificity in the diagnosis of bacterial infections.^119,120 However, subsequent preclinical experiments and clinical trials with ^99mTc-ciprofloxacin or ¹⁸F-ciprofloxacin demonstrated variable specificity and an inability to reliably differentiate infection from sterile inflammatory processes.^121–124 Development of this tracer proved controversial due to low specificity attributed to nonspecific binding to dead bacteria,¹²⁵ as well as nonspecific binding interactions with host neutrophil DNA.¹²⁶ Commercial development of ^99mTc-ciprofloxacin was suspended after an unsuccessful phase II trial in which it failed to differentiate sterile inflammation from infection.¹²⁷ Nevertheless, ^99mTc-ciprofloxacin has continued to be used by some investigators for the diagnosis of infections.^128,129 Multiple additional fluoroquinolones, including lomefloxacin,¹³⁰ fleroxacin¹³¹ and trovafloxacin,¹³² have been radiolabeled with ¹⁸F and/or ^99mTc, but preliminary studies have demonstrated inconsistent findings.^106,114 Although some have promising results for bacteria-specific imaging, extensive preclinical evaluation for the uptake mechanisms should be performed before further development in order to avoid past mistakes.

Cephalosporins

As a class of β-lactam antibiotics that directly targets the cell wall, multiple cephalosporins have been radiolabeled and evaluated for bacterial imaging. Initial experiments with ^99mTc-ceftriaxone in animal models (mice, rats, and rabbits) suggested higher accumulation in infected versus inflamed tissues.^133–135 Preliminary studies in humans with ^99mTc-ceftriaxone showed promising results with 80% sensitivity to detect orthopedic infections.¹³⁶ In 2012, Kaul et al. reported SPECT imaging with ^99mTc-ceftriaxone in 36 patients with clinical and radiological suspicion of an orthopedic infection. They reported a sensitivity and specificity of 85.2% and 77.8%, respectively, for the diagnosis of infection.¹³⁷ The high protein binding of ^99mTc-ceftriaxone (>90%) led to delayed elimination, and a high background. Additional preclinical experiments have been done with other cephalosporins such as ^99mTc-cefepime, which demonstrated selective accumulation in an E. coli-infected thigh muscle compared to a heat-killed E. coli and a turpentine oil induced inflammation in a rat model.¹³⁸ Similar experiments have been reported for ^99mTc-labeled cefazolin,¹³⁹ cefoperazone,¹⁴⁰ cefotaxime,¹⁴¹ ceftazidime,¹⁴² ceftizoxime¹⁴³ and cefuroxime.¹⁴⁴

Antifolates

Based on the structure of trimethoprim, a synthetic antibiotic that inhibits the bacterial dihydrofolate reductase, an enzyme conserved in most bacteria that is involved in DNA synthesis and the folate pathway, Sellmyer et al. developed ¹⁸F-fluoropropyl-trimethoprim (¹⁸F-FPTMP), showing in vitro accumulation in S. aureus, E. coli and P. aeruginosa.¹⁴⁵ In a mouse myositis model, ¹⁸F-FPTMP identified sites of bacterial infection with E. coli and S. aureus and differentiated infection from sterile inflammation and tumor (Fig. 6). The authors also evaluated the biodistribution of ¹⁸F-FPTMP in non-infected non-human primates with promising results for its translation to humans.

Figure 6. ¹⁸F-FPTMP PET/CT imaging of E. coli myositis.

A mouse infected live E. coli (8 log₁₀ CFU injected into the right lower leg) was imaged after injection with ¹⁸F-FPTMP, ~7.4 MBq IV. There is uptake in the infected hind limb muscle (arrow) 4 h after infection, but not in the area of turpentine injection (arrowhead). Next-day imaging with ¹⁸F-FDG, ~11.1 MBq IV, shows uptake in both infection and chemical inflammation 1 h after injection. Reprinted with permission by the National Academy of Sciences of the USA from Sellmyer et al. 2017.¹⁴⁵

Several other radiolabeled antibiotics with different mechanisms of action have been evaluated.^106,114 However, the differences between the experimental designs, makes the comparison amongst them difficult. The variability between animal models (mice, rats or rabbits), injected activities and image acquisition times are significant, and, most importantly, the number of bacteria used for inducing the infection ranges from 5 to 10 log₁₀ colony forming units (CFU).¹¹⁴ Moreover, given the rapid growth of bacteria in tissues (generation time ~20 min for most pyogenic bacteria), information on the bacterial burden at the time of imaging/radioactivity quantification rather than at the time of injection would be the critical determinant of sensitivity.

LEUKOCYTES/ANTIBODIES

In vitro leukocytes, labeled with ¹¹¹In (¹¹¹In-oxine-leukocyte) or ^99mTc (^99mTc-HMPAO-leukocyte), remains one of the most commonly clinically performed molecular imaging test for most infections in the immunocompetent population.¹⁴⁶ However, in vitro labeling of leukocytes has significant disadvantages since the labeling procedure is labor intensive, involves contact with blood products and may be difficult to perform in leukopenic patients, since it depends on intact chemotaxis, the number and types of cells labeled, and the cellular response to the specific infection.^29,147 Furthermore, since most of the radiolabeled cells are neutrophils, leukocyte imaging is much less sensitive in infections that do not elicit a predominantly neutrophilic cellular response, such as tuberculosis.¹⁴⁸

Multiple radiolabeled antibodies (e.g. ^99mTc-besilesomab, ¹¹¹In-IgG) and antibody fragments (e.g. ^99mTc-sulesomab) have been developed and evaluated with promising results.^111,149 Besilesomab, a murine-derived monoclonal antibody of the IgG1 κ isotype, radiolabeled with ^99mTc has been used with variable results in the evaluation of FUO, joint infection and osteomyelitis.^{150–152 99m}Tc-sulesomab is a 50 kDa fragment antigen-binding (Fab) portion of an IgG1 class murine monoclonal antibody that binds to the normal cross-reactive antigen-90 (NCA-90) present on leukocytes. Although sensitive for diagnosis of soft tissue infections, ^99mTc-sulesomab is not specific for bacteria.^153,154

While most of these compounds target the inflammation associated with infection, antibodies specifically targeting bacteria have also been reported, although more work is needed to improve the several pharmacokinetic challenges including slow clearance from non-target sites. Early research with a ^99mTc-labeled anti-S. aureus antibody showed encouraging results in the rabbit endocarditis model.¹⁵⁵ Using a similar endocarditis model in rats, Pinkston et al. reported a monoclonal antibody (MAb 69) against the major component of Enterococcus faecalis pili, EbpC.¹⁵⁶ After radiolabeling, ⁶⁴Cu-DOTA-MAb 69 PET imaging successfully identified E. faecalis endocarditis. Targeting a different pathogen, in 1988 Rubin et al. reported a ¹²⁵I-labeled anti-P. aeruginosa specific monoclonal antibody that was able to discriminate between P. aeruginosa and S. aureus infection sites in a rat model.¹⁵⁷ Recently, Wiehr et al. developed a ⁶⁴Cu-labeled polyclonal antibody targeting YadA, an outer membrane protein, essential for in vivo virulence of Yersinia enterocolitica.^{158 64}Cu-NODAGA-YadA PET allowed the detection of bacteria in the spleen, although a clear identification of the pathogen in the liver was not possible due to the nonspecific uptake of the radiolabeled antibody and free-copper accumulation. Using a similar approach and the similarities between M. tuberculosis and Mycobacterium bovis (Bacillus Calmette–Guérin, BCG), Malpani et al. developed a ¹³¹I-labeled anti-BCG antibody.¹⁵⁹ Localization of ¹³¹I-anti BCG was observed at the site of TB lesions in a rabbit model, but significant concentration of ¹³¹I-anti BCG was also noted in the background and major organs like heart, liver, spleen, and kidneys. Subsequently, Lee et al. reported the use of a ¹³¹I-labeled antibody against BCG,¹⁶⁰ demonstrating that the BCG-specific antibody F(ab′)₂ accumulated in TB lesions of rabbits (induced with by the inoculation of heat-killed, sonicated M. tuberculosis in muscle) while washing away from syphilitic lesions. However, it is possible that the results could be attributed to non-specific iodine uptake in the inflammation site.¹⁶¹

APTAMERS/OLIGOMERS

Aptamers are single-stranded oligonucleotides with high affinity and specificity to the target molecules, ideal characteristics for a bacteria-specific imaging tracer.¹⁶² Most of the work with aptamers for bacteria-imaging has been based on optical imaging probes.¹⁶³ In 2015, S. aureus-specific aptamers SA20, SA23 and SA34 were radiolabeled with ^99mTc and showed high uptake in the infection site compared to background.¹⁶⁴ Recently, using aptamers designed to bind to the peptidoglycan of Gram-negative and Gram-positive bacteria, Ferreira et al. recently described ^99mTc-Antibac1 and showed it could localize S. aureus infection in a mouse model although it has high binding to plasma proteins, which could limit its use.¹⁶⁵ Using a similar approach, Chen et al. synthesized a radiolabeled oligomer complementary to the bacterial 16S rRNA (^99mTc-MORF), and evaluated its accumulation in a K. pneumoniae infection in a myositis mouse model.¹⁶⁶ A more detailed review of the use of aptamers and oligomers for bacterial imaging is also available.¹⁶⁷

BACTERIOPHAGES

Bacteriophages are viruses that show no specificity for mammalian cells and infect bacteria exclusively by attaching to specific surface receptors and insert their genetic material into the cell to use it as a host for reproduction.¹⁶⁸ This approach was initially used by Rusckowski et al. who reported that the ^99mTc-labeled bacteriophage M13 had a 1.6 times higher activity in the E. coli or S. aureus-infected mouse thigh compared to inflamed muscle.¹⁶⁹ A subsequent publication from the same group described the bacteriophages P22, E79, VD-13 and 60, labeled with ^99mTc using MAG(3) as the chelator. The P. aeruginosa-specific bacteriophage ^99mTc-E79 had promising results being able to differentiate infection from inflammation in a myositis mouse model.¹⁷⁰ More recently, Cardoso et al. evaluated another ^99mTc-labeled P. aeruginosa-specific bacteriophage (PP7), describing a 1.68 times higher signal in the infected side compared with sterile inflammation.¹⁷¹ In these studies, the authors describe positive signal from the gastrointestinal tract, suggesting that the bacteriophages are binding to the host microbiome.

VITAMINS

Biotin (vitamin B7) is a precursor of acetyl CoA carboxylase in bacteria and eukaryotic cells.^{172 111}In-labeled biotin, alone or combined with streptavidin, has been reported for bacterial SPECT imaging in patients and animal models,^173–175 with promising results in the diagnosis of vertebral osteomyelitis. However, this approach is unlikely to be bacteria-specific since streptavidin accumulates at the infection site as well as the surrounding inflammation.¹⁷⁶ Additional biotin analogs have also been radiolabeled with ¹⁸F and evaluated in E. coli-infected rats.¹⁷⁷ Recently, Baldoni et al. reported an analog of cyanocobalamin (vitamin B12) labeled with ^99mTc, ^99mTc-PAMA(4)-Cbl.¹⁷⁸ This tracer had a time-dependent accumulation in vitro in S. aureus and E. coli, and ex vivo tissue biodistribution showed a higher concentration of tracer in the infected implants.

OTHER AGENTS

The negatively-charged phospholipids in bacterial membranes bind to bis(zinc(II)-dipicolylamine) (Zn-DPA). This approach was used with fluorescent Zn-DPA that accumulated in vitro in Gram-positive and Gram-negative bacteria, using optical imaging to monitor in vivo imaging of infection mouse models.^179,180 Subsequently, Liu et al. developed ¹¹¹In-DOTA-biotin and Zn-DPA-biotin, non-covalently linked by streptavidin to form the complex ¹¹¹In-DOTA-biotin-SA-Zn-DPA-biotin.¹⁸¹ Using a myositis mouse model, the signal of the ¹¹¹In-labeled Zn-DPA complex was 2.8-fold higher in the infected animals compared to inflamed controls. Using a different chelator, Rice et al. used ¹¹¹In-DTPA-Zn-DPA SPECT to successfully identify the Streptococcus pyogenes infection in the mouse model.¹⁸² However, since dead mammalian cells have an increased negative charge on the cell wall, Zn-DPA was found to also bind to necrotic and apoptotic tissues reducing its specificity for bacteria.^{183 99m}Tc and ¹⁸F-labeled analogs of Zn-DPA have been reported to successfully evaluate death and cardiomyocyte apoptosis following acute myocardial infarction.^184–186

A different approach targeting a bacterial enzyme was reported by Panizzi et al. During infection with coagulase-positive Staphylococci (e.g. S. aureus), prothrombin binds to staphylocoagulase, a fibrinogen-binding protein, to form an active complex that has fibrinogen-clotting capabilities. Using a small inhibitory peptide that binds to prothrombin for radiolabeling, PET/CT imaging with ⁶⁴Cu-DTPA-prothrombin (⁶⁴Cu-DTPA-ProT) was able to detect S. aureus valve vegetations in a mouse model of endocarditis.¹⁸⁷

CONCLUSION

For years, non-specific tracers that target infection-associated inflammation have been widely used in clinical practice. Their application has led to early detection of infectious diseases, deeper understanding of bacterial pathogenesis, antibiotic therapy decision making and treatment follow up. However, in the last decade, increasing interest and research efforts have been invested in search of specific probes for imaging bacterial infections. The ability to rapidly and non-invasively discriminate between infectious and sterile inflammation and the potential of differentiating between bacterial classes (and species) will prove to be very valuable for the practicing clinician. Although most of the reported bacteria-specific tracers are in preclinical stages, some are rapidly moving to the clinical setting with promising results. The next couple of years promise to be an exciting period for bacterial-specific imaging, as the search for the ideal tracers continues to be a challenge, but the goal does not seem impossible to achieve anymore.