*4.1. Porcine Infection Protocol*

The animal protocol was approved by the Danish Animal Experimental Board, journals no. 2012-15-2934-00123 (original approval) and 2017-15-0201-01239 (renewed approval, no substantial changes to the protocol), and all procedures followed the European Directive 2010/63/EU on the protection of animals used for scientific purposes. The protocol for the haematogenous induction of osteomyelitis in domestic pigs has been detailed elsewhere [23]; for general background, see [14–16].

Briefly, *S. aureus* was inoculated into the right hind limb of juvenile (19–25 kg) Danish Landrace–Yorkshire crossbred female pigs. The pigs were pre-acclimated for at least one week, during which they were housed in groups in boxes with restricted access to food (Dia plus FI, DLG, Copenhagen, Denmark) and *ad libitum* access to tap water. The temperature was 20–24 ◦C, the humidity was 45–65%, and there were 12 h of darkness and 12 h of light in the stables. The pigs came from a specific pathogen free (SPF) herd and were clinically examined by a veterinarian before inoculation. The pigs were fasted for approximately 16 h before premedication with midazolam and Stresnil and propofol anaesthesia. After inoculation, the pigs were individually housed. The *S. aureus* used was the porcine strain S54F9 [25], and 8000 to 30,000 CFU/kg (colony forming units per kg body weight) were inoculated. To selectively infect the right hind limb, bacteria were injected into the right femoral artery. To further reduce the possibility of systemic infection, the pigs were administered penicillin (10,000 EI/kg) at the onset of the first clinical signs of disease; for such juvenile pigs this dosage has previously been shown to allow the development of osteomyelitis while minimizing cases of systemic infection [23]. To reduce pain in the days until euthanasia, the animals were treated every 8 h with buprenorphine (Temgesic; 0.3 to 0.9 mg intramuscularly, depending on clinical symptoms). Osteomyelitis was allowed to develop for approximately one week, after which the pig was scanned with PET and computed tomography (PET/CT, see below) and euthanized. If a pig reached predefined humane endpoints [23] before this time, it was euthanized (and not scanned).

Although the protocol was originally designed for inducing osteomyelitis [17], several pigs also developed soft tissue infections in the inoculated hind limb, typically related to bone infection or the site of inoculation. This turned out to be an advantage, as it allowed us to compare osteomyelitis and soft tissue infections. Some of the pigs also developed lung abscesses. However, these abscesses were outside the field-of-view of the dynamic PET scans and are therefore not included in this kinetics study.

#### *4.2. Animals and Lesions*

This study includes eight pigs dynamically scanned with PET/CT using [68Ga]Ga-DOTA-Siglec-9 (details below). The characteristics of these pigs are summarized in Table 1. These eight pigs are a subset of the animals from the overall osteomyelitis project but represent all pigs in the project scanned with this tracer.

Before euthanasia, each pig was also PET/CT scanned with [18F]FDG; these scans have been reported earlier and found [18F]FDG to be a sensitive (but unspecific) marker of *S. aureus* infection [17,19,26]. Infectious lesions were identified on [18F]FDG PET/CT scans. As part of the post-euthanasia analysis, selected lesions were verified by necropsy to be suppurative and to be caused by the inoculated *S. aureus* strain, S54F9.

For the identified lesions, volumes of interest (VOIs) for PET data analysis were drawn on the [68Ga]Ga-DOTA-Siglec-9 PET/CT scans using Carimas 2.9 software (Turku PET Centre, www. turkupetcentre.fi/carimas). For osteomyelitis lesions, the VOI drawing was based on the CT part of the scan, while for soft tissue lesions the VOI drawing was guided by the PET scan.

For each lesion, a reference VOI was drawn in the anatomically corresponding position in the left, non-infected limb.

#### *4.3. Radiochemistry*

The radioactive labelling of DOTA-Siglec-9 has previously been discussed and described in detail [22]. The [68Ga]Ga-DOTA-Siglec-9 radiosynthesis method called method 3 in the reference was applied in this study.

In summary, 68Ga was eluted from a 68Ge/ 68Ga generator (GalliaPharm, Eckert & Ziegler, Berlin, Germany), trapped on a cation-exchange cartridge (Strata-XC 33 u Polymeric Strong, Phenomenex, Værløse, Denmark), and eluted from the cartridge with an acidified acetone solution. The pH was adjusted using HCl (0.1 M in metal-free water), and acetone was removed by heating. A solution of DOTA-Siglec-9 peptide in metal-free water was added, and 68Ga incorporation took place. Water was added to the mixture, which was then run through a preconditioned C-18 Sep-Pak cartridge to trap [ 68Ga]Ga-DOTA-Siglec-9. The product [68Ga]Ga-DOTA-Siglec-9 was released from the cartridge with ethanol and diluted with saline. After this process, the product was ready for injection.

After 25 min, a 62% non-decay-corrected (ndc) yield of the product was obtained. The [68Ga]Ga-DOTA-Siglec-9 was found by a radio HPLC system to be more than 98% radiochemically pure, and the specific radioactivity was approximately 35 MBq/nmol. Representative radio-HPLC chromatographs are shown in Supplementary Figure S11.

#### *4.4. Dynamic PET Scans*

Before scanning, each pig was anaesthetized with propofol, and catheters were implanted in the jugular vein and carotid artery [17]. After an initial CT scan, the pig was dynamically PET scanned for 60 min in 23 time frames: 8 × 15 s, 4 × 30 s, 2 × 60 s, 2 × 120 s, 4 × 300 s, and 3 × 600 s. The [68Ga]Ga-DOTA-Siglec-9 tracer was injected into the jugular vein at the start of the PET scan. The tracer activities are shown in Table 1. All of these scans were performed at the Department of Nuclear Medicine, Aalborg University Hospital.

Pigs no. 6–23 were scanned on a GE VCT Discovery 64 PET/CT scanner (GE Healthcare, Chicago, IL, USA). The scan field covered 15 cm in the axial direction and was positioned over the pelvis and the hind limbs. The images were reconstructed with an ordered subset expectation maximization (OSEM) algorithm (3D Vue Point, GE). The reconstruction parameters were 2 iterations, 28 subsets, a 128 <sup>×</sup> 128 matrix in 47 slices, a 5.5 <sup>×</sup> 5.5 <sup>×</sup> 3.3 mm3 voxel size, and a 6 mm Gaussian filter.

Due to scanner replacement, pigs no. 24–26 were scanned on a different scanner than the previous pigs. Pigs no. 24–26 were scanned on a Siemens Biograph mCT (Siemens, Erlangen, Germany) with time-of-flight (TOF) detection. The scan field covered 22 cm in the axial direction, positioned over the pelvis and the hind limbs. The images were reconstructed with an OSEM algorithm without using the resolution recovery option (setting "Iterative + TOF"). The reconstruction parameters were 3 iterations, 21 subsets, a 400 <sup>×</sup> 400 matrix in 1.02 <sup>×</sup> 1.02 <sup>×</sup> 2.03 mm<sup>3</sup> voxels, and a 3 mm Gaussian filter.

On both scanners, image reconstruction included decay-correction to the start of scanning and attenuation-correction based on the CT scan.

#### *4.5. Blood Samples*

An arterial blood sample was drawn before the tracer was injected (zero-sample). During the dynamic PET scan, 27 blood samples were drawn. All blood samples were drawn manually from the carotid artery, and the precise time (seconds) of each sample was recorded. Time zero was the time of injection, which was also the scan start time.

In pigs no. 6–10, the blood samples were drawn every 5 s for 50 s (10 samples), at 60, 80, 100, 120, and 150 s post-injection (p.i., 5 samples), and at 3, 4, 5, 6, 8, 10, 15, 20, 30, 40, 55, and 70 min p.i. (12 samples), i.e., 27 blood samples. Samples for analysis of the fraction of radioactivity originating

from the parent tracer (rather than from radioactive metabolite products or free gallium) were drawn at 2, 5, 10, 15, 25, 35, 50, and 70 min p.i.

In pigs no. 22–26, the blood sample timing was slightly optimized. Blood samples were drawn every 5 s for 40 s (8 samples), at 50, 60, 80, 100, 120, and 150 s p.i. (6 samples), and at 3, 4, 5, 6, 8, 10, 15, 20, 25, 30, 40, 50, and 60 min p.i. (13 samples), i.e., 27 blood samples. Blood samples for analysis of the parent tracer fraction were drawn at 2, 5, 10, 15, 25, 40, and 60 min p.i.

As noted in Table 1, pig no. 26 was scanned twice, and 5 mg of "cold" (unlabelled) DOTA-Siglec-9 was injected before the second scan. Blood sampling was independently performed for the two scans.

Plasma samples were obtained by collecting the supernatant after the centrifugation of whole blood samples. Aliquots of the samples were counted in a calibrated Wizard 2480 gamma counter (PerkinElmer, Turku, Finland) with an energy window from 450 to 1200 keV. The counting results were converted to decay-corrected radioactivity concentrations (Bq/mL at the time of injection).

The plasma samples for analyses of the parent tracer fraction were denatured by thoroughly mixing 0.5 mL plasma with 0.5 mL acetonitrile, after which the mix was centrifuged (approximately 1 min, 10,000 rpm) to accelerate the precipitation of the plasma proteins. An aliquot was collected for HPLC analysis (see below). The radioactivity of the precipitate was determined with the Wizard 2480 gamma counter. Protein binding was calculated as

$$\text{protein binding} = \frac{\text{precipitate radioactive} / 0.5 \text{ mL}}{\text{plasma radioactive concentration}} \times 100\% \tag{1}$$

using decay-corrected activities.

For the determination of parent tracer fractions in a sample, 0.2 mL of the supernatant was diluted with 0.8 mm water; this dilution was run through an HPLC with a fractionation collector. The fraction collector was set up with a 20 s delay to compensate for delays in the system. Twenty fractions of 45 s each were collected and counted in the Wizard 2480 gamma counter.

#### *4.6. Tissue Samples and Immunofluorescence (IF)*

As a proof of concept, the surface expression of VAP-1 in infected tissue was tested by an augmented protocol in one pig.

Approximately 10 min before euthanasia, pig no. 25 was administered 10 mg of VAP-1-binding antibody as an intravenous (i.v.) injection (10 mL injected liquid). The antibody was 1B2, a mouse IgM against human VAP-1 that also recognizes porcine VAP-1 [27]. The i.v. injection allowed 1B2 to bind to the VAP-1 expressed on cell surfaces, but not to the VAP-1 within intact cells.

After euthanasia of the pig, soft tissue samples were collected from acutely inflamed areas (phlegmon/early abscess) located peripheral to the osteomyelitis in the distal right femur and from similarly positioned non-inflamed areas in the left hind limb. All samples were embedded in a cryo-compound and frozen in petroleum spirit (VWR, Søborg, Denmark, cat. no. 87125.320) cooled with dry ice. The tissues were stored at −80 ◦C until use.

Immunofluorescence (IF) analysis was performed on 5 μm thick frozen sections of these samples. The first of two serial sections from the inflamed area (right limb) were stained with anti-VAP-1 mAb (1B2) or a class-matched negative control antibody, 7C7 (10 μg/mL; 1 h at room temperature) and then with a secondary antibody (Alexa 555-goat-anti-mouse IgM 1:100, Southern Biotech 1020-32; 30 min at room temperature), followed by nuclear staining with Hoechst 1:10,000 in PBS for 5 min, Thermo Scientific 6249, Waltham, MA, USA). The combined IF signal thus represented the VAP-1 within the cytoplasm as well as the VAP-1 expressed on cell surfaces. The second section was stained only with the secondary antibody; thus, the IF signal represented only the VAP-1 accessible to the i.v. injected antibody. A pair of sections from the left limb was similarly stained to represent the corresponding signals from non-infected tissue.

#### *4.7. Input Function and Metabolite Correction*

For each dynamic PET/CT scan, the decay-corrected plasma samples were used as a basic input function.

For metabolite correction, the parent tracer fraction data were fitted with a Hill-type function:

$$f(t) = 1 - \frac{(1-a)t^b}{c + t^b},\tag{2}$$

where *t* is the sampling time (seconds since injection). The function starts at *f*(0) = 1 (thus assuming no metabolism before injection) and has an asymptotic value f(∞) = *a*. The parameters *a*, *b* and *c* were fitted for each injection of [68Ga]Ga-DOTA-Siglec-9.

In the following text, the *uncorrected input function* denotes the total radioactivity concentration (decay-corrected Bq/mL) from plasma samples, and the *corrected input function* denotes the fraction *f*(*t*) multiplied by the uncorrected input function.

As blood samples were taken from the carotid artery (a short distance from the heart), while the PET data were acquired over the hind end of the pig (a longer distance from the heart), the PET data were delayed by some seconds relative to the blood plasma data. To correct for these delays, an offset to the plasma time stamps was determined for each pig using the method described in [19]. The largest of these corrections was 8 s.

#### *4.8. Kinetic Modelling*

Data were modelled using the three kinetic models shown in Figure 6. This was performed twice for each model: once with the uncorrected input function and once with the corrected input function. The modelling was performed using software available on the Turku PET Centre website (fit2k for 1TCM, fit3k for irr2TCM, fit4k for rev2TCM) [28].

Data points were weighted based on time frame length (*L*) and decay:

$$w = L \times \exp(-\lambda t) = L \times 0.5^{t/T} \tag{3}$$

This weighting scheme mirrors the overall count statistics of the decay-corrected PET data. Unlike weights based on counts in a VOI, the weights from Equation (3) do not in themselves contain noise. See reference [19] for a more detailed discussion of this weighting scheme.

For in vivo imaging, the volume of distribution in tissue (*VT*) is defined as the ratio of tissue concentration to input concentration at a time when a steady-state has been reached [29], i.e.,

$$Vr = \frac{\text{tissue concentration}}{\text{input concentration}} \,\left(\text{at steady-state}\right) \tag{4}$$

The measurement of tissue concentration in Bq/cm<sup>3</sup> (from the PET scan) and input concentration in Bq/mL (from plasma samples in the gamma counter) results in mL/cm3 as the unit of *VT*.

For a given model with reversible uptake, the relationship between *VT* and the model parameters can be theoretically calculated. In the cases of the 1TCM and the rev2TCM models, these relationships are [29]:

$$\text{1TCM}:\ V\_T = \frac{K\_1}{k\_2} \tag{5}$$

$$\text{rev2CTM}:\quad V\_T = \frac{K\_1}{k\_2} \cdot \left(1 + \frac{k\_3}{k\_4}\right) \tag{6}$$

For models with irreversible uptake, such as irr2TCM, a steady-state is never reached and *VT* is not defined. Mathematically, the tracer input concentration will be continually decreasing, while the uptake in the irreversible compartment will monotonically increase, and over time, the fraction in Equation (4) will diverge instead of converging.

The parent tracer fraction appeared to decrease faster in pig no. 26 (both scans) than in the other pigs, and was close to zero after approximately 30 min post-injection (see Supplementary Figure S1). The corrected input function in this case would therefore be close to zero (expectedly with high percentage errors) at late time points, which could lead to unreliable estimation of *VT* = tissue concentration/input concentration. For these reasons, modelling of pig no. 26 with the corrected input function was performed using only the data within the interval 0–30 min rather than the full interval 0–60 min.

#### *4.9. Evaluation*

In addition to visual inspections of the fits, the three models were compared using the corrected Akaike information criterion (AICc), which favours a good fit, but penalizes the use of excessive fitting parameters [30,31]. The absolute value of AICc depends on both the data and the model, but for a given data set, the lowest value of AICc indicates the most favourable model.

Furthermore, Patlak plots [32,33] were calculated to help determine whether uptake was reversible or irreversible. A linear Patlak plot (after an equilibration time) is a sign of irreversible uptake, while a system with only reversible uptake will result in a non-linear Patlak plot that eventually approaches a constant value. The Patlak plot is directly calculated from the data, without the assumption of any specific uptake model. The Patlak plots were based on the data from 10 to 60 min (10–30 min in pig no. 26 with corrected input function).

As a measure not requiring modelling, standardized uptake values (SUV) were also calculated, and the correlation between SUV and the volume of distribution *VT* was determined. The SUV calculation was based on the time interval 10 to 30 min, chosen as a time interval after the passage of the bolus peak, but not so late that a considerable part of the tracer with reversible binding (i.e., tracer not remaining in the tissue indefinitely) would have left the tissue yet.

#### *4.10. Statistics*

To compare the infected and control VOIs (i.e., right vs. left), a two-tailed paired *t*-test was used, with a significance level of *p* < 0.05. The normality of the differences was tested with the Shapiro–Wilk W test. The correlation between the volume of distribution *VT* and SUV was defined using Pearson analysis. Statistics were calculated with StatsDirect version 3.1.14 (www.statsdirect.com).

#### **5. Conclusions**

Using the VAP-1-targeted leukocyte ligand Siglec-9, the immunofluorescence analysis of infected tissue samples indicated that VAP-1 was expressed on the cell surfaces in infected tissue, while surface VAP-1 was not observed in non-infected tissue.

The uptake kinetics of [68Ga]Ga-DOTA-Siglec-9 with localized infection in pigs were found to be well described with a reversible 2-tissue compartment model, similar to the model used by Retamal et al. [12] in a study of severe lung inflammation, also in anaesthetized pigs.

We found that the parent tracer fraction decreased relatively rapidly, but despite this finding, we were unable to demonstrate an advantage of taking tracer metabolism into account in the analysis. More detailed analyses of the radioactive species occurring after the i.v. injection of [ 68Ga]Ga-DOTA-Siglec-9 in the body are warranted.

The [68Ga]Ga-DOTA-Siglec-9 uptake, evaluated as the volume of distribution, showed affinity to infection in soft tissue; however, no increased uptake in bone infections (osteomyelitis) could be demonstrated. This difference may be related to a previous report that found infected soft tissue to be more highly perfused than infected bone tissue (osteomyelitis).

**Supplementary Materials:** The following are available online, Figure S1: Parent tracer fraction curves, Figures S2 to S9: Model fits and Patlak plots, Figure S10: Volume of distribution (*VT*) as a function of SUV, Figure S11: Sample radio-HPLC graphs, Table S1: VOI volumes, Tables S2 and S3: Akaike information criterion (AICc) results

using, respectively, the uncorrected and the corrected input function, Table S4: Volume of distribution (*VT*) as determined by rev2TCM.

**Author Contributions:** Conceptualization, L.J., A.R., S.J., S.B.H., P.A., A.K.O.A., O.L.N., S.B.J.; Methodology, L.J., P.A, A.K.O.A., O.L.N., S.B.J.; Software, L.J. and V.O; Validation, L.J. and S.B.H.; Formal Analysis, L.J. and S.B.H.; Investigation, S.J., P.A., A.K.O.A., O.L.N. and S.B.J.; Resources, A.R., S.J., P.A., A.K.O.A. and S.B.J.; Data Curation, L.J. and V.O.; Writing—Original Draft Preparation, L.J.; Writing—Review & Editing, L.J., A.R., V.O., S.J., S.B.H., P.A., A.K.O.A., O.L.N., S.B.J.; Visualization, L.J. and A.R.; Supervision, P.A., O.L.N. and S.B.J.; Project Administration, P.A., A.K.O.A., O.L.N. and S.B.J.; Funding Acquisition, A.R., A.K.O.A., O.L.N. and S.B.J.

**Funding:** This work was supported by the Danish Council for Independent Research, Technology and Production Sciences, grant number 0602-01911B (11-107077), and by Jane and Aatos Erkko Foundation.

**Acknowledgments:** The authors thank Timo Kattelus for technical assistance on the graphics. Part of the material in this paper has been presented at the European Conference on Medical Physics, Copenhagen 2018 [34].

**Conflicts of Interest:** Sirpa Jalkanen owns stocks in Faron Pharmaceuticals Ltd. The remaining authors declare no conflicts of interest.

#### **References**


**Sample Availability:** Data are available from the corresponding author upon reasonable request.

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
