*3.2. The Volume of Distribution VT*

The *VT* for [68Ga]Ga-DOTA-Siglec-9 in soft tissue infections was found to be higher than that of the corresponding control sites, (red circles in Figure 10, left part). In bone infections (osteomyelitis), however, *VT* was not significantly different between the infection and control sites (blue squares in Figure 10, left part).

Retamal et al. [12] used the same tracer to study lung inflammation, also performing kinetic modelling with the model called rev2TCM in our terminology (with no mention of correction for parent tracer fraction), and found that the model described the time–activity curves well. The study also included protein binding, which was found to be constant over time, at a level of approximately 20% in healthy pigs and approximately 50% in the inflamed pigs. Our results (cf. Figure 4) are mostly in accordance with the first of these numbers, which could reflect that from a systemic perspective, a local infection in a single limb is more similar to a healthy pig than to a pig with severe inflammation in both lungs. Retamal et al. found increased uptake of [68Ga]Ga-DOTA-Siglec-9 in inflamed lungs, which is consistent with our results on uptake in soft tissue infection.

In summary, [68Ga]Ga-DOTA-Siglec-9 shows increased uptake in infected (and inflamed) soft tissue compared with control tissue; however, this study fails to demonstrate elevated uptake in infected bone (osteomyelitis). These quantitative results correspond to the visual impression of the sample image in Figure 1, where increased uptake is clear in infected soft tissue but not in the infected bone (cf. with pig no. 25 in Tables 2 and 3). Note, however, that this comparison is partly qualitative. The determination of *VT* on a reliable absolute scale will depend on improved knowledge on the nature of the radioactive metabolite products of [68Ga]Ga-DOTA-Siglec-9. The moderate correlation between SUV and *VT* indicates that the volume of distribution gives information that is not just a complicated version of the SUV.

In a previous study [18] (in part performed on the same animals), we found only a small increase of blood perfusion in osteomyelitis lesions, while blood perfusion was considerably increased in soft tissue infections. As speculated in that study, an ineffective vascular response to infection may lead to too few leukocytes reaching the infected bone, in part explaining why osteomyelitis is difficult for the body to fight. Similarly, despite the uptake of [68Ga]Ga-DOTA-Siglec-9 in infected tissue, imperfect perfusion can impair effective uptake, which may explain the difference in results for soft tissue infections and osteomyelitis.

#### *3.3. Scan after the Injection of "Cold" DOTA-Siglec-9*

Pig no. 26 was scanned twice, and the second scan was performed after the injection of 5 mg of unlabelled DOTA-Siglec-9 peptide, which was intended to block the VAP-1 receptors. The uptake curves for both scans are shown in Supplementary Figures S8 and S9. For the infected lymph node, the bolus passage and initial uptake show differences between the first and the second scan, but otherwise the two sets of curves are very similar. Quantitative distribution volumes are listed in Supplementary Table S4. Rather than the expected decrease from receptor blocking, both lesions in pig no. 26 show an approximately 20% increase in *VT* from the first scan to the second when the uncorrected input function is used. With only two lesions in one pig, however, it is difficult to draw conclusions.

Using the corrected input function, Supplementary Table S4 shows pronounced *VT* differences (still increases) for the two lesions in pig no. 26, but we are hesitant to draw conclusions from these results, as they may reflect the sensitivity of *VT* to the correction (cf. Section 3.1). In addition, pig no. 26 unfortunately showed the lowest parent tracer fractions (cf. Supplementary Figure S1) and therefore had the largest sensitivity to possible errors in the correction of the input function.

### *3.4. Limitations*

A porcine model has the advantage over, e.g., a murine model that the physical sizes involved in both surgery and scans are larger, but the disadvantage is that the cost per animal is higher. Accordingly, this study is limited by a relatively small number of animals.

The non-traumatic osteomyelitis protocol has the advantage that it is a very realistic model of haematogenous osteomyelitis (in humans and animals) and infections were reasonably limited to the right hind limb, but the numbers and locations of infections varied among the animals. However, within the individual pig, the other hind limb could be used as a control, and for soft tissue, the uptake measured as the volume of distribution (*VT*) in infected versus control sites showed quite clear differences (Figure 10). As already noted, the study does not claim to report *VT* on a robust absolute scale, which would also require data on protein binding in all animals.

A scanner replacement resulted in the last three pigs (no. 24–26) being scanned on a PET/CT scanner of a different brand and a newer generation than the initial scanner, enhancing the spatial resolution of the PET images. This allowed refined VOI drawing in these pigs, as reflected in the generally smaller VOI sizes for these animals (cf. Supplementary Table S1). More precisely drawn VOIs could reduce a possible partial volume effect, in which case uptake would be expected to be more pronounced in pigs no. 24–26 than in the pigs scanned on the initial scanner. However, the results for *VT* based on the uncorrected input function (cf. Supplementary Table S4) do not indicate any pronounced effect. Using the corrected input function, pigs no. 25 and 26 do have high *VT* values, but these values reflect the pronounced correction for the (possibly artificially low) parent tracer fractions in these pigs (cf. Supplementary Figure S1), while an effect of voxel size should also be reflected in *VT* calculated using the uncorrected input function.

Another potential source of error may be that the scans were performed on anaesthetized pigs, as the anaesthetic may affect blood flow and hence the kinetics of the tracer. However, it was not practically feasible to scan awake pigs. Propofol was chosen because it provides relatively uniform and safe anaesthesia over many hours.

The use of penicillin in the animals may have reduced the extent of the studied infections, but has previously been found to provide a better balance between the successful development of osteomyelitis and the avoidance of systemic infections requiring euthanasia of the animal [23]. The use of opioids can have immunosuppressant effects, but so can pain reactions, which are reduced by the pain-killer effect; buprenorphine was chosen because it has shown weaker immune effects than morphine and fentanyl [24]. As the target molecule VAP-1 for the tracer is not related to the infecting agent but is instead part of the natural immune system response, there is no reason to expect direct interference between [68Ga]Ga-DOTA-Siglec-9 and the antibiotic pharmaceuticals (penicillin and buprenorphine).

#### **4. Materials and Methods**
