*2.8. Correlation Between Standardized Uptake Value (SUV) and VT*

Only *VT* values based on the uncorrected input functions were considered. The Pearson correlation coefficient was *r* = 0.54 (*r*<sup>2</sup> = 0.29, *n* = 48, *p* < 0.0001), indicating a moderate correlation. A plot is shown in Supplementary Figure S10.

Often, SUV is not evaluated as an absolute value but relative to a reference tissue. If relative rather than absolute values of *VT* and SUV were plotted, then correlation rose to *r* = 0.66 (*r*<sup>2</sup> = 0.43, *n* = 24 pairs, *p* = 0.0005). A plot is shown in Figure 11.

**Figure 11.** Correlation between relative distribution volume and relative SUV. In the plot *VT-*ratio = *VT,*infected / *VT,*control and SUVratio = SUVinfected / SUVcontrol, where for each infection VOI, the control VOI is at the corresponding location in the non-infected limb of the same animal VOI (see main text).

#### **3. Discussion**

In this study, we investigated the uptake of [68Ga]Ga-DOTA-Siglec-9 in juvenile pigs with localized *S. aureus* infections, including both bone infection (osteomyelitis) and soft tissue infection.

As in previous studies [10,13], the tracer was found to have an affinity for infected tissues. Siglec-9 is known to bind to VAP-1, and IF staining was consistent with the expression of VAP-1 on cell surfaces in the infected soft tissue.

Kinetic modelling showed that [68Ga]Ga-DOTA-Siglec-9 had reversible uptake kinetics, which could be described with a two-tissue compartment model (i.e., rev2TCM from Figure 6). Visually, the models with fewer parameters provided good fits in many cases (see the 1TCM and irr2TCM curves in Figures 7 and 8); however, the 1TCM curves provided poor fits for the initial part of many curves, and the Patlak plots revealed that models with irreversible uptake such as irr2TCM could not be generally applied (cf. Figure 9). Likewise, the analysis by the AICc values (cf. Supplementary Tables S2 and S3) favoured rev2TCM in the large majority of cases.

Based on these results, the following discussion will assume rev2TCM as the chosen model. Physiologically, rev2TCM indicates that the tracer is taken up by the tissue (first compartment), and then the tracer binds to receptors or otherwise changes status in the tissue (second compartment), but with the possibility of unbinding/changing back (reversible model).

#### *3.1. Using the Corrected or the Uncorrected Input Function*

The radio-HPLC analysis of blood samples revealed a rapid decrease in the parent tracer fraction (Figure 5). However, modelling showed no clear advantage of using a corrected input function. In most fits, the visual difference between using the uncorrected or corrected input function was only minor (compare Figures 7 and 8).

Physiologically, fitting with the uncorrected input function assumes that any radioactive metabolites have the same uptake kinetics as the parent tracer. In contrast, fitting with the corrected input function assumes that the radioactive metabolites have no uptake at all. The real situation is probably somewhere in between.

Mathematically, the inclusion of such a correction has the disadvantage of including a source of uncertainty, which becomes especially important in a case such as the present where the correction in the late part of the scans was considerable (cf. Figure 5).

The volume of distribution (*VT*) measures how concentrated the tracer is in tissue relative to plasma (input), cf. Equation (4) in the Material and Methods section. As shown by comparing the scales in Figure 10, the calculated values of *VT* depend markedly on the selection of the uncorrected or corrected input function. This dependency is a simple consequence of the math. The corrected input function is by definition only some fraction of the total, uncorrected input function (cf. Figure 5). A given radioactivity concentration in the tissue will be relatively higher when compared to a low number (the corrected input function) than when compared to a higher number (the uncorrected input function).

In summary, our data showed a rapid decrease in the parent tracer (cf. Figure 5), but unfortunately the modelling did not allow us to distinguish between the PET signal from the parent tracer (i.e., [68Ga]Ga-DOTA-Siglec-9) and the PET signal from other radioactive species formed in vivo (i.e., all metabolites containing the 68Ga isotope, possibly including free 68Ga).

Consequently we reason that use of the corrected input function will lead to increased uncertainty of *VT* due to possible errors in the measurement of parent tracer fraction. When the uncorrected input function is used, this pitfall is avoided, although at the cost of risking over-simplification. Pragmatically, the following discussion will therefore focus on the results found with the uncorrected input functions, bearing in mind that the physiological reality is likely more complicated than the model. Further studies investigating the nature of the radioactive products of [68Ga]Ga-DOTA-Siglec-9 formed in vivo are warranted.
