¹⁸ F-Fluoride positron emission tomography/computed tomography for noninvasive in vivo quantification of pathophysiological bone metabolism in experimental murine arthritis

DBA/1 mice were bred at the animal facility of the Jena University Hospital (Jena, Germany). All animal studies were approved by the local commission for animal protection (Thüringer Landesamt für Verbraucherschutz, Bad Langensalza, Germany; registered number 02-045/08). Arthritis was induced as described elsewhere[13]. In brief, DBA/1 mice were immunized subcutaneously with 400 μg of recombinant human G6PI in emulsified complete Freund’s adjuvant (Sigma-Aldrich, Taufkirchen, Germany). Macroscopic evaluation of arthritis was performed according to severity of clinical manifestations in wrist and ankle joints, in metacarpophalangeal and metatarsophalangeal joints, and in digits and toes. Swelling and redness in wrist and ankle joints and in metacarpophalangeal and metatarsophalangeal joints, respectively, were graded from 0 to 3. A score of 0 indicates no macroscopically recognizable signs of arthritis, 1 indicates swelling and redness, 2 means strong swelling and redness and 3 indicates massive swelling and redness. Additionally, the number of digits and toes with inflamed joints were divided in half to avoid assessment imbalance because inflammation in the G6PI-induced arthritis model is located mainly in proximal joints of the paws[2]. To calculate the total clinical score per animal, results from all paws were summed. For PET/CT measurements, mice were anesthetized with 1.5% to 2% isoflurane (Deltaselect, Dreieich, Germany) vaporized in oxygen (1.5 L/min) to prevent animal movement and reduce imaging artifacts. Respiration of mice under anesthesia was monitored. ¹⁸ F-fluoride (half-life = 109 minutes; Eckert & Ziegler, Bad Berka, Germany) with an activity of 10.4 ± 0.8 MBq was injected intravenously into the lateral tail vein. Longitudinal arthritis imaging was performed at various time points of acute and chronic clinical arthritis (n = 3 to 6 mice per time point). We obtained dynamic PET scans for kinetic analysis of tracer uptake in nonimmunized mice (n = 3).

In vivo¹⁸ F-NaF imaging was performed using a Siemens Inveon small-animal multimodality PET/CT system (Preclinical Solutions, Siemens Healthcare Molecular Imaging, Knoxville, TN, USA), characterized by the combination of two independently operating PET and CT scanners. Radial, tangential and axial resolutions at the center of the field of view of the PET module are 1.5 mm for this imaging modality[14, 15]. PET image acquisition was carried out with a coincidence timing window of 3.4 ns and an energy window of 350 to 650 keV. PET data acquisition was performed for 1,800 seconds, starting 50 minutes after tracer application. In kinetic analysis, PET data acquisition lasted 5,400 seconds and started concomitantly with radiotracer injection. Images were reconstructed into three-dimensional images using Fourier rebinning and three-dimensional ordered-subset expectation maximization algorithm. Attenuation of PET data was corrected based on the CT measurements. The CT module consists of a cone-beam X-ray μCT source (50-μm focal spot size) and a 3,072 × 2,048–pixel X-ray detector. In our μCT imaging protocol, we used an axial–transaxial resolution of 2,048 × 2,048–pixel, 80 kVp at 500 μA, 360° rotation and 360 projections per bed position for paw measurements and 3,072 × 2,048–pixel, 220° rotation and 120 projections per bed position for whole-animal attenuation scans, respectively. CT images were reconstructed using a Shepp-Logan filter and cone-beam–filtered back projection. To reduce stress due to overly prolonged measurement times, only high-resolution μCT data from hindpaws were acquired for correlation analysis of bone surface assessment with ¹⁸ F-fluoride PET and CT.

Quantitative analysis of ¹⁸ F-fluoride accumulation in the period from 50 to 80 minutes after ¹⁸ F-fluoride injection or 0 to 90 minutes, respectively, was enabled by image fusion technology in Siemens Inveon Research Workplace 4.0 software. For dynamic PET analysis, scans were started 5 seconds before radiotracer injection and continued for 90 minutes. The 90-minute data set was divided into 45 time frames (6 × 10 seconds, 6 × 20 seconds, 7 × 60 seconds, 10 × 120 seconds, 10 × 180 seconds and 6 × 300 seconds) during histogramming to construct time–activity curves. For static analysis, data from 75 to 80 minutes after ¹⁸ F-fluoride injection were analyzed in a single time frame. Radioisotope activity in the venous blood pool or in fore- or hindpaws, reflecting bone incorporation of ¹⁸ F-fluoride, was measured as standardized uptake value (SUV; g/ml) using PMOD 3.15 software (PMOD Technologies Ltd, Zurich, Switzerland) or Siemens Inveon Research Workplace 4.0 software. Guided by maximum intensity projection images, volumes of interest were drawn as spheres (ellipsoids) over anatomic structures of bones and joints. A threshold of 40% (max–min) was used to separate the site of tracer accumulation from background tissue signaling. Visualization of skeletal elements of the hindpaws was also done using PMOD 3.15 software.

Analyses of metatarsal CT data were performed using Definiens Developer XD™ 2.0.3 build 2015 software (Definiens, Munich, Germany). For segmentation of metatarsal bones, the automatic threshold routine was used. Then, individual bones were discriminated by watershed segmentation. Oversegmented fragments were merged manually to achieve three-dimensional objects of each bone. After identification of metatarsals 1 to 5, a local threshold was applied on each metatarsal bone to refine bone margins and remove remaining nonbone pixels. The threshold was calculated as the difference between mean pixel gray value and standard deviation of pixel gray values. Quantitative measurements of bone surface and bone volume were calculated for each metatarsal bone.

Statistical differences between groups were evaluated using the nonparametric Mann-Whitney U test. Correlation analyses (Spearman test) were performed for PET data and clinical scoring or PET and CT data, respectively, whereas CT surface data of all metatarsals, per paw, were summed. Statistical significance was accepted at P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001). In bar charts and text, data are given as arithmetic means and standard errors of the mean (SEM). All calculations were performed using the software package IBM SPSS Statistics version 20.0 (IBM, Ehningen, Germany).

Immunization of animals with G6PI induced severe paw inflammation in DBA/1 mice (Figure 1A). Experimental disease was characterized by redness and swelling in wrist and ankle joints, as well as in dorsal areas of the paws and in metatarsophalangeal, metacarpophalangeal and phalangeal joints. Macroscopic scoring revealed onset of polyarthritis around 9 days after disease induction, maximum of inflammation at day 14 (d14) and a subsequent decrease until d33 (Figure 1B). Histopathology revealed bone erosion in skeletal elements of the inflamed paws (Figure 1C).

Radiotracer applied intravenously in the tail vein was immediately (0 to 10 seconds postinjection) transported to the heart via the blood flow (Figure 2A). Subsequent whole-body distribution of ¹⁸ F-fluoride resulted in thoracic and intestinal PET signaling (10 to 20 seconds postinjection). After 3 minutes, the pattern of ¹⁸ F-fluoride PET signaling indicated onset of bone accumulation, and signaling from the kidneys and bladder revealed partial radiotracer excretion. Twenty minutes after radiotracer injection, signaling of nonexcreted ¹⁸ F-fluoride was restricted mainly to bone and continued to increase until the end of measurement at 90 minutes.

Concomitantly, the quantitative time–activity curve analysis of ¹⁸ F-fluoride paw uptake revealed rapid radiotracer accumulation beginning immediately after injection and tracer signaling remaining at almost the same level from 40 to 90 minutes postinjection (Figure 2B). At 45 minutes postinjection, 90% of ¹⁸ F-fluoride uptake was achieved (SUV_mean = 1.8), compared to bone radiotracer signaling after 90 minutes (SUV_mean = 1.9). In contrast, blood pool tracer activity measured in the venae cavae region showed a rapid decline after an activity maximum (SUV_max = 15.2), within minutes following injection, reflecting rapid clearance of ¹⁸ F-fluoride from the blood.

Detailed three-dimensional anatomical μCT measurements allowed identification of metatarsal and phalangeal bones and joints, and PET imaging revealed radiotracer uptake in the paws during the course of experimental arthritis (Figure 3A). Low ¹⁸ F-fluoride accumulation before immunization switched to a distinct tracer uptake in the acute (d14) and chronic (d28) stages of disease and declined during late remitting arthritis (d50). Coregistration of PET and CT data enabled exact localization of ¹⁸ F-fluoride accumulation in arthritic mice.Detailed, three-dimensional anatomical information obtained by high-resolution μCT enabled visualization of the bone surfaces of phalangeal, metatarsal, tarsal and tibial bones. Compared to nonarthritic control animals (Figure 3B, left), pathologic bone metabolism in the mice with G6PI-induced arthritis induced a distinct increase in bone roughness (Figure 3B, right).

To analyze the usability of the ¹⁸ F-fluoride PET imaging modality for quantification of pathological bone metabolism in experimental arthritis models, we measured ¹⁸ F-fluoride uptake in all paws of mice with G6PI-induced arthritis at five different time points of disease. Before onset of clinical arthritis (that is, in bone tissue without pathological changes (d8), there were only low levels of ¹⁸ F-fluoride accumulation in wrist (Figure 4A) and ankle (Figure 4B) joints, measured as SUV. In acute clinical arthritis at d11, paw inflammation was associated with an increase in bone metabolism and induced a significant increase of ¹⁸ F-fluoride uptake compared to d8 (P = 0.019 (wrist) and P = 0.002 (ankle)). Mean SUVs at d8 were 2.08 ± 0.1 in carpal joints and 4.0 ± 0.2 in tarsal joints, and at d11 mean SUVs were 3.8 ± 0.6 in carpal joints and 5.8 ± 0.5 in tarsal joints. Mean ¹⁸ F-fluoride uptake was further significantly increased at d18 (4.4 ± 0.5 (carpal) and 6.3 ± 0.5 (tarsal)) and d25 (5.6 ± 0.6 (carpal) and 6.5 ± 0.6 (tarsal)), followed by a decrease at d39 (4.4 ± 0.5 (carpal) and 5.0 ± 0.4 (tarsal)). PET data revealed a 1.3- to 2.8-fold increase in ¹⁸ F-fluoride bone uptake in the paws at progressive stages of inflammatory arthritis, which always differed significantly from ¹⁸ F-fluoride accumulation before onset of clinical disease. This finding indicated a rise of pathological bone-degrading and bone-forming processes leading to joint dysfunction, characteristic for RA and its animal models.

To examine the general effects of arthritis on ¹⁸ F-fluoride uptake in elements of the skeleton, we quantified PET signaling in medial and distal femoral bones. SUVs revealed high tracer uptake (SUV > 10) in trabecular distal areas of the femur (Figure 4C). Except for late remitting arthritis (d39), SUVs remained at a similar level at different stages of G6PI-induced arthritis, indicating only a slight influence of pathological bone metabolism associated with arthritis on distal femoral tracer uptake. In contrast to trabecular distal areas, mean SUVs in median cortical areas of the femur ranged from 2.1 to 2.2 in all stages of arthritis (Figure 4D), additionally negating general effects of arthritis on bone accumulation of ¹⁸ F-fluoride at sites without clinical manifestations of disease. ¹⁸ F-fluoride PET/CT measurements in healthy mice without any pathological conditions revealed a nonequal pattern of bone tracer uptake (Additional file1: Figures S1A and S1B). There were hotspots of PET signaling in proximal joints of the pectoral girdle; in the pelvis, including the knees; and in spinal and cranial bones. Volume-rendering of the knee joints revealed a coincidence of increased ¹⁸ F-fluoride accumulation and trabecular bone (Additional file1: Figure S1C). Therefore, we assume that the increased surface in trabecular bones, meeting the demands of distinct mechanical stress or saving weight, is the cause for the observed heterogeneous pattern of ¹⁸ F-fluoride uptake in skeletal bone.

To validate assessment of pathophysiological bone metabolism by ¹⁸ F-fluoride PET imaging, bone parameters in μCT data were quantified. Bone surface showed no difference for metatarsals 1 to 5 between unimmunized mice and mice with acute inflammatory arthritis (Figure 5A, d14). In chronic and remitting arthritis (Figure 5A, d28, d35 and d50), bone surface was significantly increased due to bone lesions or dysfunctional bone formation. The course of bone surface assessment was consistent with the course of SUV data obtained by ¹⁸ F-fluoride PET. The volume of metatarsal bones revealed a slight increase at d35, followed by a subsequent decline to previous levels in late experimental arthritis at d50 (Figure 5B). Overall, bone volume was largely unaffected by arthritis pathogenesis.

To validate quantification of pathological bone metabolism by ¹⁸ F-fluoride PET imaging in experimental arthritis, we performed correlation analysis of quantitative PET results with semiquantitative clinical scoring data and assessment of bone surface roughness destruction by μCT. Although inflammation and bone destruction are associated but functionally not necessarily constrained mechanisms of arthritis pathophysiology, macroscopic clinical scoring and SUV of ¹⁸ F-fluoride were significantly correlated (Figure 6A). More importantly, there was a significant correlation of quantitative assessment of pathological bone metabolism by ¹⁸ F-fluoride SUV and assessment of bone surface by μCT (Figure 6B), demonstrating the feasibility of these modalities for quantification of bone destruction in experimental arthritis. Therefore, small-animal ¹⁸ F-fluoride PET/CT is a reliable imaging method for in vivo quantification of arthritis-induced bone erosion and bone malformation, correlating with assessment of pathophysiologic bone surface alterations using μCT.