Synthesis of [¹⁸F]FDG was by standard nucleophilic substitution using an IBA Synthera module (IBA, Louvaine-la-Neuve, Belgium). [¹⁸F]FDS was synthesized following previously described methods, with some modifications (Weinstein et al., 2014; Li et al., 2017). Briefly, [¹⁸F]FDG was reduced with 2 mg sodium borohydride at 35°C for 30 min with periodic shaking, the reaction was stopped by the addition of 1.4 ml of a sodium acetate/HCl solution (0.9/0.4 M), and the final pH was adjusted to 7 with NaOH. The final product was purified using an Alumina N Light Sep-Pack cartridge and sterile filtered (0.22 μm). Radiochemical purity was >97%, as determined by radio-thin layer chromatography (radio-TLC).

The available synthesis methods of [¹⁸F]FPABA (Zhang et al., 2018; Li et al., 2020) are complex and require a precursor that is difficult to prepare. Consequently, we prepared a new precursor [methyl 4-N,N-di(Boc)-2-nitro-4-aminobenzoate (2)] from a commercially available reagent to enable a one-step nucleophilic substitution synthesis in an automated synthesis module (Figure 1). To a solution of methyl 2-nitro-4-aminobenzoate (1) (0.25 g, 34 mmol) and di-tert-butyl dicarbonate (7.4 g, 68 mmol) in methylene chloride (10 ml), cooled with ice water, was added one equivalent of 4-(dimethylamino)pyridine. After stirring for 30 min at room temperature (RT), the mixture was diluted with an additional 50 ml of methylene chloride, washed with brine, and dried over MgSO₄. Removal of the solvent gave 370 mg of the crude product, which was purified by silica gel column chromatography using CH₂Cl₂/methanol (95/5) as eluent to yield 290 mg (58%) of 2 as a white solid. For the radiosynthesis, [¹⁸F]F⁻ produced by irradiation of H₂¹⁸O in a 18/9 cyclotron (IBA, Belgium) was trapped on a QMA cartridge and eluted with 0.5 ml of K₂CO₃/Kryptofix 2.2.2. After azeotropic drying, 2 mg of precursor 2 was added to the reactor in 1 ml N, N-Dimethylformamide, heated to 150°C for 20 min, hydrolyzed at 110°C for 10 min with 0.5 ml of 5 M NaOH, and then cooled to 40°C and neutralized with 1.5 ml of 1.7 M HCl. The crude product was purified by radio-HPLC using a VP 250/16 Nucleosil 100–7 C18 column, and 90/10 trifluoroacetic acid (TFA) 0.1% in H₂O/MeCN as the mobile phase. [¹⁸F]FPABA (eluted at 9–10 min) was collected in 40 ml of water, reformulated by trapping on a MCXplus cartridge, and eluted with 2 ml of ethanol/NaOH 1 M (50/50) into a citrate buffer solution to a final volume of 5 ml. All radiosynthesis steps took place in an automated synthesis module (AiO36, Trasis, Belgium), for which a specific program and user interface (Figure 2) were developed.

Radiochemical purity was determined by HPLC on a Mediterranea Sea C18 (150 × 4.6) column with a 0.1% TFA/MeCN gradient at 1 ml/min. [¹⁸F]FPABA was eluted at 5.7 min, whereas the free [¹⁸F]F⁻ was eluted at 2 min.

The following bacterial and fungal reference strains (ATCC) were used: Staphylococcus epidermidis ATCC 12228 and ATCC 35984, S. aureus ATCC 29213 and 25923, Cutibacterium acnes ATCC 11827, Escherichia coli ATCC 25922, and Pseudomonas aeruginosa ATCC 27853 and Candida albicans ATCC 10231. Selected clinical isolates from patients with prosthetic material were also included (Supplementary Table S1).

All microorganisms, except C. acnes, were cultured the night before on solid agar, followed by 20–24 h of culture in tryptic soy broth (TSB, bioMérieux) at 37°C under constant agitation. Cutibacterium acnes was cultured for 48 h on solid agar, then for another 48 h in thioglycolate broth (TG, bioMérieux) under anaerobic conditions. Broth medium was selected to achieve better tracer uptake compared to a solid media. In addition, a prolonged incubation time was set to allow the microorganisms to consume most of the D-glucose.

The next day, the incubation broth was diluted in 4 ml 0.9% NaCl and adjusted with more incubation broth or saline until an optical density of 0.75 McFarland units was reached. Dilution in saline was chosen to reduce the amount of D-glucose in case any traces remained that might be available and compete with the carbohydrate tracers ([¹⁸F]FDG and [¹⁸F]FDS). Three replicates (500 μl) of each microorganism were incubated with radiotracers (100 μl, 37 ± 11 MBq/ml) for 2 h under agitation, pelleted by centrifugation, and washed with PBS. Final radioactivity in all pellets was measured using a gamma counter (Hidex Automatic Gamma Counter) calibrated for fluorine-18 and normalized to the initial number of colonies (Bq/10⁶ CFU: colony-forming units). Aliquots of E. coli ATCC 25922 with the same optical density were used as negative control after inactivation at 90°C for 30 min. Counts per minute were corrected for background and decay. The experiment was repeated three times on each strain in three independent experiments, using triplicate of PBS as background.

The number of viable microorganisms was quantified at the beginning and end of each experiment. CFU per ml (CFU/ml) values were determined by dilutions and plating on trypticase soy agar (TSA) under optimal growth conditions depending on the microorganism. Initial CFUs were measured to normalize to activity (Bq/10⁶ CFU) and as a control. The number of colonies in the final pellet was measured to monitor the growth of the microorganisms.

All procedures involving animals were carried out in accordance with the guidelines of the European Communities Council Directive (2010/63/EU) and the Spanish Government (RD 53/2013) and were approved by the Animal Experimentation Ethics Committee of the University of Navarra (Protocol no. 103-17). Four-week-old male and female ICR mice (n = 42; males = 16 and females = 26) were purchased from Envigo and socially housed in the animal facilities of the University of Navarra under controlled conditions (22 ± 2°C, 12-h light/12-h dark cycle; relative humidity 55 ± 10%).

Staphylococcus aureus ATCC 29213 and E. coli ATCC 25922 were selected as representative GP and GN bacteria to develop the mouse model of acute myositis.

Fresh bacterial strains were incubated in TSB broth overnight. Inoculated bacteria were diluted and incubated for an extra hour to obtain exponential growth. Bacterial load was adjusted to 7–8 log₁₀ CFU. 50 μl of live bacteria were injected into the right hindlimb of the animals; the same volume of heat-inactivated inoculum (90°C, 30 min) was injected into the left hindlimb to simulate sterile inflammation. The mice were imaged with [¹⁸F]FDG, [¹⁸F]FDS, or [¹⁸F]FPABA 16 to 18 h after the initial infection for comparison.

For data analysis, the Mann–Whitney U test was used to compare two groups, and the Kruskal Wallis test for multigroup comparison. Multigroup mean comparisons of in vitro growth at the beginning and the end of experiments (CFU/g) were analyzed with Sidak’s test. Data are shown as mean ± SD.

The [¹⁸F]FDS synthesis procedure yielded a high purity product (>97%) over 20 different syntheses. The most suitable TLC chromatography method for analysis was the one used for [¹⁸F]FDG following European Pharmacopeia instructions (01/2014:1325). The final product always appeared as a sharp peak at a retention factor (Rf) of 0.15–0.20, while the precursor appeared at over 0.5. We checked that the final pH of the formulation was 7.0 ± 0.5 and adjusted with the corresponding acid or base if necessary.

The precursor, methyl 4-N,N-di(Boc)-2-nitro-4-aminobenzoate (used for radiosynthesis of [¹⁸F]FPABA), was prepared in a one-step reaction with an average yield of 58%. The implementation of radiosynthesis in the automated system (TRASIS AllinOne 36 module) proved to be reliable, reproducible, and amenable to GMP implementation. Overall, radiosynthesis took less than 80 min (including radio-HPLC purification and reformulation) and [¹⁸F]FPABA was produced with a radiochemical yield of 6% (final activities were in the range of 1.3–1.8 GBq in 4 ml) and > 95% radiochemical purity, as assessed by radio-HPLC (Supplementary Figure S2). [¹⁸F]FPABA obtained a specific activity of 395.53 ± 189.07 GBq/μmol at the end of synthesis. The injectable solution of [¹⁸F]FPABA had a pH of 6, contained 10% ethanol as a stabilizing agent and was stable for at least 8 h as determined by radio-HPLC.

All bacteria except Pseudomonas aeruginosa and Stenotrophomonas maltophilia incorporated [¹⁸F]FDG (Figure 3A). [¹⁸F]FDG uptake in GP bacteria was higher than in GN bacteria. Among Gram-positives, Streptococcus agalactiae obtained the highest uptake. All yeasts tested, C. albicans ATCC 10231 and clinical strain (CS 2502) accumulated much more [¹⁸F]FDG than bacteria.

Enterobacterales (i.e., E. coli and Enterobacter) was the only group of bacteria that accumulated [¹⁸F]FDS (Figure 3B). The uptake values of [¹⁸F]FDS were at least three times lower than those of [¹⁸F]FDG, for example, in the reference bacteria E. coli ATCC 25922 ([¹⁸F]FDG = 605.28 ± 91.91 Bq/10⁶ CFU and [¹⁸F]FDS = 160.23 ± 71.76 Bq/10⁶ CFU).

All bacteria and yeasts accumulated [¹⁸F]FPABA (Figure 3C). There were no clear differences between GPs, GNs, yeasts, and the microaerophilic GP, C. acnes. The two highest accumulations were obtained in S. aureus CS 16557 and E. coli ATCC 25922. Comparison of different strains of the same species of S. aureus showed large differences (S. aureus CS 3082 = 954.29 ± 568.26 Bq/10⁶ CFU and S. aureus CS MR-methicillin resistant 16557 = 5,192 ± 1334.49 Bq/10⁶ CFU). Regardless of their type, the heat-killed bacteria showed virtually no uptake of the radiotracers (Figure 3).

Only some yeast species accumulated [¹⁸F]FDS. Clinical and reference isolates of C. albicans took up the radiotracer, whereas C. glabrata did not (Figure 4A). Candida albicans showed more accumulation of radiolabeled monosaccharides ([¹⁸F]FDG and [¹⁸F]FDS) than bacteria. In contrast, [¹⁸F]FPABA accumulation was higher in E. coli than in fungi (Figure 4B). Interestingly, [¹⁸F]FPABA was the only radiotracer that was taken up by P. aeruginosa (Figure 4C).

The variation in the initial and final number of viable cells (CFU/ml) is shown in Supplementary Table S3. We found no relevant variation for most of the microorganisms. However, [¹⁸F]FDG resulted in a significant decrease in the growth of C. acnes (p ≤ 0.0001). In addition, C. acnes ATCC 11827, P. aeruginosa and S. agalactiae showed a non-statistically significant decrease in growth with [¹⁸F]FDG. Incubation with [¹⁸F]FDS induced a statistically significant reduction of CFU/mL in S. agalactiae, C. acnes, and P. aeruginosa. Only E. coli growth appeared to increase with incubation with [¹⁸F]FDS (p = 0.0229). [¹⁸F]FPABA did not change the number of bacteria or yeast in these experiments, with the sole exception of the clinical strain of E. coli.

The [¹⁸F]FDG, [¹⁸F]FDS, and [¹⁸F]FPABA PET images detected E. coli myositis infection (Figure 5A; Supplementary Figure S4). The SUVr values were similar for [¹⁸F]FDG (1.55 ± 0.39), [¹⁸F]FDS (2.09 ± 0.84), and [¹⁸F]FPABA (1.92 ± 1.19) with no statistically significant differences (Figure 5B).

The post-mortem  measurements of radioactivity in infected and inflamed tissues, presented as %ID/g, was not significant for [¹⁸F]FDS (Figure 5C). By contrast, [¹⁸F]FPABA readily discriminated between infection (%ID/g = 0.089 ± 0.078) and inflammation (%ID/g = 0.008 ± 0.002) and showed a statistically significant difference (p = 0.0079; Figure 5D).

Mean bacterial load counts at the infection site were similar, approximately 7 log₁₀ across all mice and probe groups (Supplementary Table S5). In addition, strong signals with minimal background were detected at the infection site with bacterial loads of 5.59 log₁₀ CFU/g with [¹⁸F]FDG, 5.5 log₁₀ CFU/g with [¹⁸F]FDS, and 6.17 log₁₀ CFU/g with [¹⁸F]FPABA.

[¹⁸F]FPABA and [¹⁸F]FDG detected S. aureus infection in the right limb, as shown in Figure 6A; Supplementary Figure S4. A non-specific [¹⁸F]FDG signal was observed in the inflamed muscle. As expected, the [¹⁸F]FDS PET images showed no uptake in the inflamed and in the infected muscle. Radiotracer uptake, expressed as SUVr, indicated that [¹⁸F]FPABA could distinguish infection from inflammation (SUVr = 3.090 ± 0.807), but with no statistically significant differences when compared with [¹⁸F]FDG (p = 0.1763; Figure 6B).

Analysis of post-mortem biopsies demonstrated a statistically significant difference between infection and inflammation with [¹⁸F]FPABA (p = 0.0079; Figure 6D).

The means of bacterial load in all experiments are shown in Supplementary Table S5. [¹⁸F]FDG PET and [¹⁸F]FPABA detected and localized S. aureus infection with a minimum bacterial load of 6.31 log₁₀ and 7.15 log₁₀ CFU/g, respectively.

Our in vitro results expanded the number of bacteria studied (Figure 3A) compared to previous studies (Heuker et al., 2017). [¹⁸F]FDG uptake was higher in GPs than GNs in our model. Streptococcus agalactiae achieved the highest bacterial uptake, as was the case with Streptococcus pyogenes in other studies (Heuker et al., 2017). Since we observed almost no uptake in non-lactose fermenting gram-negative bacilli (P. aeruginosa and S. maltophilia), infections with these bacteria could be difficult to diagnose with [¹⁸F]FDG PET (Figure 4C). Some infectious diseases guidelines propose [¹⁸F]FDG PET imaging as a criterion for diagnosis (Habib et al., 2015; Haidar and Singh, 2022; Lauri et al., 2022) which may be underestimated if P. aeruginosa or S. maltophilia is the etiology of infection. These findings should be taken into account when interpreting PET images in biofilm-related infections (endocarditis associated with prosthetic valves or pacemakers). Mature biofilms are formed by stationary (dormant) bacteria that produce a minor inflammatory response. Pseudomonas aeruginosa is a well-known biofilm former, especially on medical surfaces, cystic fibrosis patients, or chronic wound infection. The involvement of this bacteria could reduce the negative predictive value of nuclear molecular techniques using [¹⁸F]FDG. There is also insufficient evidence on the sensitivity of [¹⁸F]FDG in non-lactose fermenters, since only case reports have reported on the use of [¹⁸F]FDG-PET as a diagnostic tool in P. aeruginosa infections (Kawamura et al., 2021). Although GP bacteria are the main biofilm producers, our data show that [¹⁸F]FDG uptake is higher in these microorganisms. This could increase the sensitivity for [¹⁸F]FDG imaging in infection related to biofilm-producing GP bacteria.

[¹⁸F]FDG accumulation in Cutibacterium acnes, which is microaerophilic and typically slow growing, was just 2 h (Figure 3A), 3 h less than the mean growth time known for this bacterium (5.1 h; Hall et al., 1994) The final CFU/ml count in the in vitro experiments showed a statistically significant decrease in the clinical strain of C. acnes and a non-significant decrease in C. acnes ATCC 11827. This could be explained by the fact that uptake of [¹⁸F]FDG by these bacteria does not require exponential growth or a suitable environment.

All yeasts tested showed higher uptake of [¹⁸F]FDG as compared to bacteria (Figure 3A). [¹⁸F]FDG could therefore be a good radiotracer for yeast infections, with the possibility of extending it to filamentous fungi. One study on the in vitro uptake of [¹⁸F]FDG showed detection of A. fumigatus, but unlike our experiments with yeasts, this uptake appeared to be lower than in bacteria (i.e., E. coli; Lai et al., 2022).

[¹⁸F]FDG uptake was increased in myositis infection, with nonspecific uptake in [¹⁸F]FDG-PET imaging of inflamed muscle tissue, notably in heat-killed E. coli myositis (Figures 4A, 5A). Toxin production, probably due to S. aureus toxins or E. coli endotoxins, induces elevated inflammation that generates high levels of glucose transporter expression in infection when compared with inflammation with heat-killed bacteria; this effect corrects SUV ratio values.

[¹⁸F]FDS was taken up in vitro by Enterobacterales species (i.e., E. coli and E. ludwigii), but not by GPs and non-lactose fermenters (Figure 3B), as described in other studies (Weinstein et al., 2014; Ordonez et al., 2017). Escherichia coli has a generation time of 84 min using ᴅ-sorbitol (Lengeler, 1975), which allowed these isolates to grow in our experiments by metabolizing [¹⁸F]FDS as sorbitol-6-phosphate would, generating fructose-6-phosphate in the process. Only E. coli ATCC 25922 showed a statistically significant increase in numbers using [¹⁸F]FDS (Supplementary Table S3). Future studies will be necessary therefore to confirm whether this bacterium shows growth with fluorinated sorbitol.

We also demonstrated uptake in yeasts such as Candida albicans (Figure 4A). The higher [¹⁸F]FDG uptake (3-fold) compared to [¹⁸F]FDS may be related to the fact that sorbitol is not the preferred carbon source in Enterobacterales and C. albicans (Figure 4B). In recent experiments, E. coli and A. fumigatus had a higher uptake (approximately 10-fold) with [¹⁸F]FDG than with [¹⁸F]FDS (Lai et al., 2022). A clear difference in uptake of [¹⁸F]FDS by Candida glabrata was shown (66.54 ± 224.67 Bq/10⁶ CFU) when compared with the other two strains of C. albicans analyzed (ATCC 10231 = 21496.06 ± 10684.79; clinical strain = 5846.24 ± 924.90 Bq/10⁶ CFU; Figure 4A). [¹⁸F]FDS is therefore useful to discriminate between yeasts with positive sorbitol assimilation (i.e., C. albicans) and those unable to assimilate sorbitol (i.e., C. glabrata; Kurtzman and Fell, 1998). This result could improve the diagnosis of disseminated/invasive Candida albicans infections (especially in the liver, CNS, eye, and other sterile locations) using [¹⁸F]FDS-PET. We obtained higher uptakes (Bq/10⁶ CFU) in C. albicans than E. coli ATCC 29213, as was the case in a recently published study (Kim et al., 2022). In another study however, the results showed that [³H]sorbitol uptake was 10-fold lower in C. albicans than in E. coli at 120 min (Lai et al., 2022). This difference in uptake probably depends on the units used for normalization. A universal indicator to normalize both bacterial and fungal biomass is challenging because of differences in growth and cell size (fungi are much larger than bacterial cells). In vitro uptake may be underestimated when normalized to the amount of protein (Kim et al., 2022). This discrepancy could probably be resolved in further experiments, by testing in vivo uptake in an animal model of acute mixed infections with C. albicans and E. coli (normalized to final CFU/g per tissue).

The [¹⁸F]FDS PET images clearly differentiated between inflammation and E. coli myositis infection although postmortem biodistribution studies found no uptake (%ID/g) differences between hindlimbs (Figures 5A,C). This diagnostic imaging with class-specific bacterial and yeast probes could be of interest when clinicians have a definite etiology of the infection. Even so, it is difficult to differentiate between all types of bacteria and microorganisms with probes because of the large number of different species. [¹⁸F]FDS is a GN-specific probe with a low limit of detection (5.5 log₁₀ CFU/g) and so can diagnose chronic infections with Enterobacterales. Its previous translation to humans (Yao et al., 2016; Zhu et al., 2016; Ordonez et al., 2021) supports safe use of this technology in chronic infections. The limitation of our results is that [¹⁸F]FDS may not differentiate between bacteria and yeasts and so adequate clinical information is needed. It would be interesting to develop animal models (with yeasts or mixed infections with bacteria and yeasts) to check and compare [¹⁸F]FDS-PET signals. To avoid negative results in specific detection of Enterobacterales and C. albicans, sequential use of other radiotracers could be considered to broaden the range of pathogens identified. In addition, [¹⁸F]FDS imaging could improve the monitoring of prolonged treatment in deep-seated Enterobacterales infections, as antimicrobial therapy may fail due to the impaired drug penetration or acquisition of antimicrobial resistance.

The proposed synthesis pathway to obtain [¹⁸F]FPABA offers a promising solution due to its various advantages, which include the ease of obtaining the radiosynthetic precursor, the simple intermediate steps using everyday laboratory reagents, all highly soluble in water. The simplicity of the pathway also facilitates synthesis on different automated radiosynthesizers if a semipreparative radio-HPLC system is available. Further improvements in synthesis, yet unexplored, may be able to increase the yield while maintaining the radiochemical purity of the final product above 97%. [¹⁸F]FPABA appears to be a better radiotracer than [¹¹C]PABA, not only because of the longer half-life of the radionuclide (110 min for fluorine-18 vs. just 20 min for carbon-11), but probably also because of the in vivo metabolism of the radiotracer, which could complicate image interpretation in the carbon-11 tracer. The accumulation of sulfonamide-like chemicals in the cell depends to a significant extent on the degree of ionization in the cytoplasm and the surrounding medium (Zarfl et al., 2008). Zarfl et al. point out that no accumulation occurs in the cell if the external pH exceeds the intracellular pH. A lower pH amplifies microorganisms that can accumulate because most bacteria have an intracellular pH of around 7 (E. coli = 7.6).

In vitro uptake of [¹⁸F]FPABA was tested in all bacterial and yeast strains (Figure 3C). [¹⁸F]FPABA appears to be a better probe for all bacteria and yeasts for diagnosis of infection, with higher in vitro uptake than [¹⁸F]FDS. Previous in vitro studies have examined S. aureus and E. coli with [¹⁸F]FPABA and [¹¹C]PABA (Zhang et al., 2018; Ordonez et al., 2022). Pseudomonas aeruginosa and Mycobacterium tuberculosis were assessed only with ³H (Ordonez et al., 2017), but no other species were tested. We obtained the same promising results in Enterobacterales, S. aureus, and P. aeruginosa. New data were obtained for C. acnes, yeast, S. epidermidis, and Enterobacterales. Unlike [¹⁸F]FDG and [¹⁸F]FDS, P. aeruginosa incorporates [¹⁸F]FPABA (Figure 4C), which makes this probe of interest in nosocomial infections. Although the innovative ⁶⁸Ga- and [¹⁸F]FIAU-radiolabeled siderophores showed potential in models for bacteria-specific imaging with P. aeruginosa (Petrik et al., 2021), they fell short when translated to human clinical trials (Zhu et al., 2016; Cho et al., 2020).

Different in vitro uptakes were obtained in strains of the same species (S. aureus and S. epidermidis), as was found in other studies performed with [¹⁸F]FPABA and S. aureus (Zhang et al., 2018). DHPS has different expression levels in the same bacterial species, which may explain the different results observed (Wang et al., 2021). Alterations in this enzyme, one of the mechanisms of bacterial resistance, correlate with the low effectiveness of sulfonamides in some species or strains. These resistance mechanisms or concomitant treatment prior to image acquisition may affect [¹⁸F]FPABA activity. In addition, AbgT transporters are exporters of PABA and interspecies variation may have an impact (Delmar and Yu, 2016). Furthermore, bacteria are capable of de novo PABA synthesis from chorismate and glutamine (Dosselaere and Vanderleyden, 2001). Therefore, future studies should evaluate possible signal variation in PABA PET imaging.

To the best of our knowledge, this is the first study of in vitro uptake of [¹⁸F]FDG, [¹⁸F]FDS, and [¹⁸F]FPABA in yeasts. Our results showed [¹⁸F]FPABA uptake in the yeasts tested and suggest that it is a promising future tool in fungal diagnostics. We suspect that these results of in vitro uptake are because DHFR (dihydrofolate reductase), an enzyme essential for folate-dependent pathways in fungi, is a valid antifungal target in C. albicans (DeJarnette et al., 2020). The DHFR enzyme is the step subsequent to PABA incorporation into DHPS. In addition, DHPS inhibitors such as sulfamethoxazole are an effective treatment for Pneumocystis pneumonia (caused by the fungus Pneumocystis jirovecii). Although folate biosynthesis pathways and enzymes have been characterized in both bacteria and plants, they are not well studied or characterized in yeasts. Our results confirm that yeast requires in vitro uptake of PABA and that [¹⁸F]FPABA could be used as a probe for yeasts in PET imaging.

[¹⁸F]FPABA PET, performed on mouse models of E. coli and S. aureus myositis, clearly differentiated between inflammation and infection and showed no uptake in the inflamed limb (Figures 5A, 6A). The SUVr values of [¹⁸F]FPABA were lower than expected (Figures 5B, 6B) because the bone showed uptake and was difficult to avoid when VOIs were drawn on the image. In post-mortem biodistribution studies, Zhang et al. observed that [¹⁸F]FPABA accumulated in the tibia at different rates (at 30, 60, and 120 min) with high ID%/g values, which might explain why some of the bone was seen on images and the SUV was less sensitive (Zhang et al., 2018). Our interpretation however was limited to visual analysis and current advances in artificial intelligence can certainly improve our image data (Schwenck et al., 2022). However, the post-mortem analysis between hindlimbs with infection (S. aureus and E. coli) and inflammation was statistically significant, and clearly demonstrated detection of infection (Figures 5C, 6C).

The mean bacterial load in acute S. aureus and E. coli myositis in our experiments (Supplementary Table S5) was around 7 log₁₀ CFU/g, and [¹⁸F]FPABA localized all these infections. This bacterial load is low compared to acute infections, where the CFU/mL is 8 log₁₀ or more (Davey and Barza, 1987; König et al., 1998). However, the infection load for chronic or bloodstream infections is even lower (Kang et al., 2014). [¹⁸F]FPABA uptake remains high regardless of the growth phase (Zhang et al., 2018) and testing this probe on lower bacterial loads in chronic infections and implant-associated biofilms could be of interest in the future. PET imaging with [¹¹C]PABA on a rabbit model of S. aureus prosthetic implant infection obtained a good discrimination ratio (target-to-nontarget tissue) of 3.17 (Ordonez et al., 2022). The main problem in that study was that the image was obtained 7 days after infection and with bacterial loads of 7-log₁₀ CFU, similar to acute infection (without the decreased sensitivity associated with low bacterial load). [¹⁸F]FPABA uptake into other typical implant-associated microorganisms, such as C. acnes, was in vitro in our study and it would be interesting to test them on in vivo models. It could be an improvement over microbiological tests that are long (14 days) and require complicated incubation atmospheres (microaerophilic/anaerobic; Schäfer et al., 2008).