The choice of fluorinated d-AAs to investigate was determined by a two factors: (1) feasibility of radiosynthesis from [18F]fluoride and (2), likelihood that the d-AA would be incorporated in the bacterial peptidoglycan cell wall, to create the desired contrast between bacteria and human inflammatory cells. 18F-Fluorinated AAs can be difficult to access by nucleophilic 18F-fluorination radiochemistry [35] and we therefore selected two AA candidates with established synthetic routes. The l-tyrosine (Tyr) derivative l-[18F]FET is clinically advanced and is a useful tool for predicting and monitoring treatment response in patients with glioma [36]. The methionine (Met) analogue, [18F]FPHCys (as both l- and d-enantiomers) has been evaluated before in cancer imaging [33]. Neither of the d-enantiomers has been evaluated previously for bacterial infection imaging. During the course of this work, d-[11C]Met was reported as a promising candidate for bacterial infection imaging, and thus evaluation of its fluorinated analogue d-[18F]FPHCys was of interest [27]. d-[11C]Alanine (d-[11C]Ala) was also described recently, showing good uptake across a panel of bacteria, [29, 37] but although accessible from [18F]fluoride, we discounted 3-[18F]fluoro-d-Ala because the L-enantiomer rapidly defluorinates in vivo [38]. Few other AAs are readily prepared by late-stage radiofluorination.

Exogenous d-AAs are appended to the muropeptides of peptidoglycan by transpeptidase enzymes [39]. Both d-Tyr and d-Met are substrates of these enzymes, which are known to tolerate a variety of unnatural side chains [40,41,42,43]. Modification of d-Met and d-Tyr with fluorine-18 was therefore reasonable.

Radiosyntheses of both d-[18F]FET and d-[18F]FPHCys were achieved as one-pot, two-stage reactions that are amenable to automation, which would be essential for future clinical translation. Furthermore, in both cases, the desired chiral centre was already established in the precursor, and a challenging chiral induction or enantiomer separation were not required in the radiolabelling procedure.

In vitro uptake studies confirmed that both d-[18F]FET and d-[18F]FPHCys became associated with live bacteria but not with heat-killed bacteria, indicating that detection of active infection was feasible. The in vitro uptake of d-[18F]FPHCys in S. aureus was higher than observed by Stewart for d-[11C]Met (ca. 20 Bq/106 cells) after 2 h [44], and similar as percentage uptake (ca. 2%) to that observed by Neumann for d-[14C]Met, although different S. aureus strains were used (Xen29 vs ATCC 12600, although Xen29 is a derivative of ATCC 12600) [27, 45, Fig S5 ESM]. Since d-[18F]FPHCys is an unnatural analogue of the parent d-Met, we were encouraged that its activity in vitro showed at least comparable uptake to the parent under similar assay conditions. Our studies revealed approximately twofold higher-uptake for d-[18F]FPHCys than d-[18F]FET in both S. aureus and P. aeruginosa. For this reason, we selected of d-[18F]FPHCys for in vivo studies.

d-[18F]FPHCys was next evaluated in vivo. Injection of the Cytodex beads and S. aureus for the infection group allowed us to assess a clinically relevant site of co-existing infection and inflammation. S. aureus is the most commonly found pathogen in SSI and accounted for around 18% of all SSI in an EU-wide study [8, 46, 47]. In addition to its prevalence in SSI, S. aureus is a clinically challenging bacterium which has increasing resistance to antimicrobials, making it a key pathogen for focus in research [48]. The bioluminescent S. aureus allowed visualisation of metabolically active S. aureus in the animals prior to the biodistribution study. The bioluminescent signal is proportional to the number of metabolically active bacteria present (although factors such as hypoxia and depth effects in imaging prevent accurate calculations of bacterial number in vivo from this data). IHC analysis of the infected and inflamed skin sections (after radioactive decay) confirmed leucocyte infiltration in both infection and inflammation sites. By day 4, S. aureus infection would have become established and akin to a clinical infection of a foreign body, and is likely to have begun biofilm formation.

In the biodistribution study, d-[18F]FPHCys showed increased uptake in S. aureus infection site versus inflammation site, which was statistically significant. Although direct comparison of d-[18F]FPHCys uptake with d-[11C]Met, d-[11C]Ala and [18F]CF3-d-Ala would be informative, this is challenging due to the differing characteristics of the animal models used here and in previously reported studies [27, 29, 31]. The ex vivo analysis of radiotracer uptake (as percent injected dose per gram) in the infection site appears generally lower for d-[18F]FPHCys than for other reported d-AAs. A comparison of the performance of d-AAs in vivo is provided in the ESM (Table S3). A number of reasons may account for this. First, the S. aureus infection model used in the present study mimics an established infection, which has been allowed to develop over 4 days. S. aureus infections are known to form biofilms [49], potentially creating a barrier to uptake of a blood circulating radiotracer compared with an acute S. aureus infection, in which bacteria were injected a few hours before an in vivo study. Second, the quantity of bacteria present in the model infection site is an important factor. Typically, acute soft tissue infections are associated with bacterial burden of 108 CFU/mL, but it has been suggested that 105 CFU/mL is a promising threshold for imaging chronic or partially treated infection [7]. Our initial inoculant contained 1 × 105 CFU per infection site, although the burden in the mice at day 4 was not determined. Both these features of our model provide clinically relevant and challenging conditions for a potential radiotracer, as would be faced in a clinical scenario. Future evaluation of d-[18F]FPHCys in additional animal models of infection, as well as reducing bacterial load further would be informative.

There was no significant difference in [18F]FDG uptake in sites of S. aureus infection compared with sterile inflammation in our animal model. This was expected because [18F]FDG uptake for infection imaging primarily represents increased glycolytic activity of inflammatory cells, although [18F]FDG does accumulate in bacteria [10, 50]. This assertion was borne out in this study by the inability of [18F]FDG to distinguish the S. aureus infection from inflammation, and supports the conclusion that d-[18F]FPHCys is specifically targeting the bacteria. Furthermore, [18F]FDG uptake in the inflammation and infection sites in our animal model (ca. 2 %ID/g) was lower than that observed for [18F]FDG in the murine myositis model used for evaluation of other d-AA radiotracers (ca. 4 %ID/g) [27, 29, Table S3]. This supports our conclusion that the lower accumulation of d-[18F]FPHCys is, at least in part, likely a result of different characteristics of the animal model.

A potential limitation of d-[18F]FPHCys, like other radiolabelled d-AAs, is that its uptake requires the bacteria to be in a metabolically active state. For all radiotracers targeting bacterial metabolism, a challenge remains for imaging infection sites that contain populations of quiescent cells, such as during antibiotic treatment. For imaging in these scenarios, radiotracers with uptake that is independent of growth phase may have an advantage. However, drug resistant strains can still be usefully visualised [22, 29].

In vivo metabolism of new radiotracers is also a key consideration. d-AAs are possible substrates of d-amino acid oxidase (DAAO), a flavoprotein that catalyses oxidative deamination of neutral d-AAs to form α-keto acids [26]. Although we did not test directly for DAAO metabolism in this study, previous studies in mice showed high metabolic stability of d-[18F]FPHCys in vivo [33]. One of the advantages of d-CF3-Ala is its stability against DAAO, and defluorination (unlike the mono-fluoro derivative, 3-[18F]fluoro-d-Ala) [31, 38].

Although S. aureus and P. aeruginosa are common culprits for SSI, d-[18F]FPHCys has not been evaluated across a full panel of pathogens, limiting the scope of this study.

d-[11C]Met has recently been evaluated in a first-in-human PET/MR study in healthy volunteers and patients with suspected joint infections [28]. While the results are promising in terms of both a favourable safety profile and an increase of d-[11C]Met uptake in suspected infections (although a gold standard for confirmed infection was lacking), an 18F-labelled analogue would have practical advantages for future application of d-AA imaging in patients with suspected infection. In our study, d-[18F]FPHCys distinguished S. aureus infection from sterile inflammation in a clinically relevant mouse model, paving the way for fluorinated d-AA imaging of S. aureus infections.