Can a glycosylated FDG tracer selectively image bacterial infection with PET?

Why standard FDG blurs infection and inflammation

The clinical problem is not detecting metabolic activity but attributing it. 18F-fluorodeoxyglucose (FDG), the widely used glucose analogue, accumulates in any metabolically active tissue, including immune cells, so it lights up the host inflammatory response as well as the bacteria. The paper states that most clinical imaging, including FDG-PET, 111In-labelled white blood cell SPECT and 67Ga citrate SPECT, detects the host response rather than the pathogen, which limits its ability to separate bacterial infection from sterile processes such as osteoarthritis or autoimmune inflammation.

The authors' proposed fix is to give the tracer a route of entry that only bacteria possess. Mammalian cells take up glucose through GLUT transporters moving sugar down a concentration gradient. Bacteria also use the phosphotransferase system (PTS), which chemically modifies the sugar as it is transported. The paper's design goal was a glucose-like molecule that the PTS accepts but mammalian GLUT and SGLT transporters largely ignore.

One-step glycosylation built a screenable tracer library

Rather than designing a single molecule, the team generated a library. O-glycosylation of clinical [18F]FDG with alcohols under acid catalysis produced alkyl glucosides in a single step, with yields of 33–92%. That range is the chemical yield of the radiosynthesis step, not a measure of bacterial uptake. The point of the library was to allow rapid screening of many candidate sugars for bacterial accumulation.

Screening in vitro identified a methanol-derived candidate, Me-2-[18F]FDG, as the strongest accumulator in Staphylococcus aureus. The mixed α,β preparation was then resolved, and the β-anomer proved to be the active component. The β-anomer could be made from a β-methyl mannose triflate precursor using the standard [18F]FDG production workflow, which the authors present as a practical advantage because it requires no modification to existing equipment.

Uptake depends on the glucose-specific PTS transporter

The mechanistic evidence comes from bacterial lysates and mutants. When β-Me-2-[18F]FDG was incubated with S. aureus lysate, the parent peak disappeared and a metabolite peak appeared that co-eluted with a 19F standard for the likely 6-phosphorylated adduct. That is consistent with PTS-style transport, in which the sugar is phosphorylated as it enters the cell.

Genetic evidence points to the same transporter. In transposon mutants of S. aureus lacking components of the glucose-specific PTS, uptake of β-Me-2-[18F]FDG was reduced relative to the parent JE2 strain. Competition assays with non-radioactive carbohydrates further supported a specific transport route rather than passive accumulation. The authors also report that the glucose-specific PTS transporter sequence is highly homologous across Staphylococcus species, which they use to explain why uptake was seen across multiple staphylococcal strains.

The selectivity claim rests on the other side of the comparison: mammalian cells. In the reported in vitro comparisons, β-Me-2-[18F]FDG uptake was high in S. aureus and low in mammalian cell lines, whereas the comparator tracer [18F]FDS showed a different pattern. The paper's interpretation is that modifications at the C1 and C2 positions make the molecule a poor substrate for mammalian GLUT/SGLT transporters while remaining acceptable to the bacterial PTS.

The β-anomer is stable and clears through the kidneys

An imaging tracer is only as useful as its background. In uninfected mice, dynamic PET/CT over 90 minutes showed rapid uptake in lung, liver and kidneys followed by bladder accumulation, indicating urinary excretion as the primary route. The estimated distribution half-life was 0.48 minutes and the elimination half-life 44.34 minutes.

Anomer choice mattered for stability. HPLC of urine collected 90 minutes after injection showed no visible [18F]FDG peak for β-Me-2-[18F]FDG, whereas small amounts of [18F]FDG appeared in urine from the mixed α,β and ethyl analogues, indicating partial degradation in vivo. The mixed preparations also showed higher heart and brain signal than the pure β-anomer. The authors attribute the difference to greater in vivo stability of β-glycosidic linkages.

Human dosimetry was extrapolated from mouse data using ICRP60 and ICRP103 models, not measured in people. The estimated effective doses were 0.026 mSv MBq−1 (male) and 0.041 mSv MBq−1 (female) under ICRP60, and 0.020 and 0.028 mSv MBq−1 under ICRP103, with the urinary bladder receiving the highest absorbed dose.

Infected tissue outshines sterile inflammation in mice

The central in vivo comparison is between live bacteria and everything else. In murine myositis models, β-Me-2-[18F]FDG accumulated at sites inoculated with live S. aureus, S. epidermidis, S. lugdunensis, S. saprophyticus, S. agalactiae or S. pyogenes, with no detectable uptake at sites inoculated with the corresponding heat-killed strain. The abstract reports more than sixfold higher signal at live gram-positive infection sites than in non-infected tissue.

Specificity was tested against inflammation without bacteria. In a lipopolysaccharide model of systemic but sterile inflammation, β-Me-2-[18F]FDG images resembled those of uninfected mice, with low background and renal elimination. By contrast, [18F]FDG showed elevated uptake in several tissues including brain and spinal cord in that model, and substantial whole-body background in the S. aureus myositis model.

Signal strength varied by infection site. In a wound infection model, β-Me-2-[18F]FDG gave nearly 17-fold higher signal in infected than uninfected tissue. In a pulmonary infection model, uptake was 11.3 ± 2.9 %IDcc−1 in infected lungs versus 0.6 ± 0.2 %IDcc−1 in uninfected lungs, alongside a measured bacterial burden of 8.9 ± 0.2 log10 CFU. In a rat vertebral discitis-osteomyelitis model, signal persisted at the inoculated level on days 2, 4 and 6. These are different models with different denominators, so the fold-changes should not be treated as one interchangeable sensitivity figure.

Tracer signal falls when oxacillin works

The therapeutic test used a myositis model with oxacillin-susceptible S. aureus (MIC 0.25 μg ml−1) in one deltoid and oxacillin-resistant S. aureus (MIC > 32 μg ml−1) in the other. Mice received oxacillin at 50 mg kg−1 as 10 injections (4 intraperitoneal, 6 intravenous) over 72 hours. Static PET scans were taken 3 hours after inoculation, before treatment, and 8 hours after the final dose.

Signal dropped after treatment at the susceptible site but not at the resistant site, and ex vivo CFU counts after imaging were consistent with that split. This is the paper's strongest functional claim: the tracer reports on bacterial viability rather than merely on tissue inflammation, because a sterile inflammatory response to dying bacteria would not be expected to disappear in step with the CFU measurement. The comparison is internal to the same animals, which strengthens it, though the sample was nine mice.

A gram-positive partner rather than a universal tracer

The paper positions β-Me-2-[18F]FDG as complementary to, not a replacement for, [18F]FDS. The authors describe the two tracers as having orthogonal bacterial sensitivity: β-Me-2-[18F]FDG is sensitive to gram-positive organisms, especially S. aureus, whereas [18F]FDS targets Enterobacterales. In their proposed clinical logic, tracer choice would follow the likely causative organism, guided by history, examination, routine microbiology and laboratory parameters. They suggest β-Me-2-[18F]FDG where Staphylococcus infection is likely, and [18F]FDS for urinary tract and hepatobiliary infections, which they state are most frequently caused by gram-negative species.

That scenario split is the paper's own proposal, not a demonstrated clinical algorithm. The evidence behind it is in vitro uptake across staphylococcal species and murine infection models; the authors note that in vitro uptake was relatively lower in Staphylococcus caprae and Staphylococcus capitis than in other species, although most strains still showed substantial uptake. No human imaging data are reported. The authors state they are preparing an investigational new drug application, including toxicity testing of the 19F standard, to support first-in-human studies.

What the mouse data do and do not settle

The most decision-relevant result is the pairing of two comparisons: high target-to-non-target contrast against heat-killed bacteria, and low signal in sterile inflammation. Together they support the claim that the tracer's accumulation tracks bacterial metabolism rather than host inflammation, which is exactly what conventional FDG-PET cannot do. The oxacillin experiment extends that claim from detection to monitoring.

The boundaries are equally specific. All infection results come from murine or rat models, with small group sizes (for example, four mice per group in several imaging comparisons and nine mice in the treatment study). Human dosimetry is extrapolated from mouse biodistribution. The tracer's validated range in this paper is gram-positive infection, chiefly Staphylococcus, so the gram-negative scenario belongs to the comparator tracer rather than to this one. Within those limits, the paper's priority is clear: the β-anomer, not the mixed α,β preparation, is the component worth developing, because it combines bacterial PTS-dependent uptake with in vivo stability and low mammalian background.

Sources used in this answer