Animals were housed in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. All experimental procedures were approved by the National Institute of Allergy and Infectious Diseases, Division of Intramural Research, Animal Care and Use Committee and were in compliance with the Animal Welfare Act regulations, Public Health Service policy, and the Guide for the Care and Use of Laboratory Animals recommendations.
Subjects
Six male rhesus macaques housed in biosafety level 3 containment, weighing 3 to 4 kg, were infected intravenously with 5 × 107 plaque forming units of monkeypox virus (MPXV Zaire 79 strain [V-79-I-005]) (for virus preparation and inoculation procedures, see Additional file 1). Three animals were treated intravenously with cidofovir (5 mg/ml/kg in Dulbecco's modified Eagle's medium; Gilead Sciences, Foster City, CA, USA) that has been shown to protect against monkeypox virus infection. The antiviral agent, cidofovir, was administered on day -1 prior to monkeypox virus challenge and on days +1, +3, +5, +7, +10, and +13 after challenge. NHPs received 25 mg/kg of probenecid by gavage 1 h before cidofovir injection to prevent cidofovir nephrotoxicity. Three animals comprised the untreated control group.
Data acquisition
Up to nine imaging sessions were conducted in each of the six animals following the procedures described previously [1]. Briefly, imaging data were acquired in animals anesthetized with isoflurane (2% to 2.5%) (Piramal Critical Care, Orchard Park, NY, USA) using a microPET scanner Focus-220 (Siemens AG, Malvern, PA, USA). This scanner has a bore size of 22 cm with an axial field-of-view of 7.6 cm and a transverse field-of-view of 19 cm [16]. Multiple static PET scans were initiated 1 h after the intravenous 18F-FDG injection (9.25 MBq/kg) and continued for 10 min for each of two bed positions on different days over 1.5 months. Three scans were performed prior to monkeypox virus inoculation (days -20, -15, and -5) and up to six scans were conducted postinoculation (days +1 or +2, +3 or +4, +7 or +8, +10, +16, and +21). The scans were conducted in the morning; the animals were fasted overnight for 12 h prior to the scanning session. The blood glucose concentrations were measured prior to the 18F-FDG injection before each scanning sessions. PET images were acquired in three-dimensional (3D) mode and reconstructed iteratively using 3D-ordered subsets expectation maximization algorithm with two iterations and nine subsets followed by 18 iterations of maximum a posteriori reconstruction [17]. Maximum a posteriori parameters were adjusted to provide a uniform spatial resolution of 1.8 mm (FWHM = 1.8 mm) in all three directions. Methods for scatter, decay, random, and attenuation correction were applied during the process of PET image reconstruction.
CT images were acquired with a CereTom® (NeuroLogica Corp., Danvers, MA, USA) 8-slice mobile head-and-neck CT scanner that was installed in close proximity to the microPET scanner. The CereTom® CT scanner provided 190 slices with 0.49 × 0.49 mm in-plane resolution and 1.25-mm slice thickness that were acquired at 120 kVp and 5 mA. CT scans were taken either immediately before or after PET imaging to ensure consistent animal position and fusion of the PET and CT scans for data analysis. Incorporating the use of the same table for both scanners eliminated the need for animal repositioning. To restrict animal motion, the animal was secured by anchoring the limbs and by controlling the level of anesthesia. CT scans were used for attenuation correction and coregistration with PET images to define anatomical localization of the LNs of interest. In addition, CT images were used to obtain the LN volume applied for SUVmean computation to determine interscan data reproducibility prior to viral challenge.
Image analysis
Data analysis was performed using MIM workstation software version 5.2.2 (MIM Software Inc., Cleveland, OH, USA). The largest LN in the axillary fossa was chosen for analysis. This LN usually is positioned close to the body surface and easy to identify. VOIs were first specified on CT images coregistered with PET images. The volume of a whole LN was delineated by manually drawing the peripheral boundary on each of the slices where it appeared (Figure 1a (i)). SUVs from all the voxels inside this volume were averaged to calculate SUVmean. The SUV with the highest value within the LN volume identified by the MIM program was recorded as SUVmax.
To compute the SUVthreshold, a `threshold' option in MIM was selected to place a sphere around the whole LN (Figure 1a (ii)). The tool averaged the voxels inside the sphere above a specified threshold that was set at 50% of the maximal voxel value. As the LN metabolic activity was often below that of surrounding tissue at baseline (Additional file 2) and during early infection (day 3/4, Additional files 2 and 3, Figure 1b), we also used a modified SUVthreshold (mSUVthreshold) method to exclude the voxels outside the LN. mSUVthreshold was calculated by averaging all the voxels with SUVs above 50% of the maximal value within a sphere inside the LN edges (Figure 1a (iii)). The diameter of a sphere was chosen based on the smallest LN axis defined on CT images. First, to specify the smallest axis, the CT images in 3D view were examined to identify the edges and the center of a LN (Figure 1c). When the approximate center of a LN was located, two perpendicular lines were drawn on LN slices in transaxial, coronal, and sagittal views following the largest and the smallest LN axis on each view (Figure 1c). When the sphere diameter was specified, the sphere was placed strictly within the edges of the LN (Figure 1a (iii)).
Additional file 2:Elevated18F-FDG uptake in tissue surrounding the LN on day 3 after virus inoculation. Maximum intensity projection (MIP) movie and representative 18F-FDG-PET images fused with CT images of an axillary LN in sagittal view acquired -5 days before and day +3 and 10 after virus inoculation. High 18F-FDG uptake is observed in tissue surrounding the LN on day +3 after virus inoculation. (MP4 4 MB)
Additional file 3:Elevated18F-FDG uptake in axillary LN of an animal that eventually became moribund. MIP movie and representative 18F-FDG-PET images fused with CT images of an axillary LN in sagittal view acquired before and on day 3 after monkey virus inoculation showing elevated 18F-FDG uptake in axillary LN of an animal that eventually became moribund. (MP4 5 MB)
We also applied a fixed dimension method, SUVfixed volume, by creating a template of three identical small spheres (0.2 cm diameter, total of 21 full voxels in all three spheres), placed contiguously within the longest axis of the LN. The sphere diameter was chosen based on the smallest LN axis (range 0.25 to 0.39 cm) among all animals determined on baseline CT images. The three spheres were transferred to each new data set by determining the center of the LN (defined by the intersection of axes in 3D view, Figure 1c), placing the middle sphere in the LN center and the other two spheres adjacent to the first one along the long axis of a LN in transaxial view. By using three small spherical VOIs, we adjusted the VOI location to the shape of the LN of each subject. Although the size and shape of the LNs differed, the VOI covered similar locations in the middle of each LN. The SUVfixed volume was computed by averaging the SUVs from 21 voxels covered by three spheres.
For SUV calculation, the radioactivity concentration from the VOI on the PET image was divided by the injected dose and normalized to the body weight of the animal and radioactive decay for the time point of 18F-FDG injection
(1)
Statistical analysis
The correlation between the volume measurement on pre-inoculation CT scans 1 and 2, 1 and 3, and 2 and 3 was calculated using the Pearson product-moment correlation coefficient (r). Application of the Kolmogorov-Smirnov test [18] confirmed that the difference between the pairs of scans followed a Gaussian distribution. To investigate the interscan reproducibility of three baseline scans, we compared scans 1 and 2, 1 and 3, and 2 and 3 for LN volumes and SUVmean, SUVmax, SUVthreshold, mSUVthreshold, and SUVfixed volume using Bland-Altman analysis [19]. The mean difference, standard deviation of the mean differences (SD), coefficient of repeatability (CR), and limits of agreement (LoA) were calculated and represented as Bland-Altman plots. The SD was calculated by squaring all the differences, adding them up, dividing them by the number of measurements, and taking the square root. The LoA were calculated by adding (upper limit) or subtracting (lower limit) the CR, defined as CR = 1.96 × SD, from the mean difference. This analysis of data reproducibility was performed with the assumption that the animal health status did not change during the time the scans were obtained 5 to 21 days prior to monkeypox virus inoculation. Unchanged animal health status was confirmed by physical examination.
Bland-Altman analysis was subsequently used to investigate the agreement between five VOI metrics for peak 18F-FDG uptake in the LNs of survivors on day 10 after inoculation. The differences obtained for each animal were plotted against the mean differences of the respective pairs of VOI measures. For acceptable agreement, the 95% LoA (±1.96 SD of the mean difference) should include 95% of the difference between the methods of measurement. Two-way repeated measures analysis of variance (ANOVA) was employed to explore the difference between treated and untreated groups or surviving and moribund groups in LN 18F-FDG uptake using SUVs with five different VOI metrics. We used 18F-FDG uptake value at different time points (days -1, -2, -3 pre- and days +1 or +2 and +3 or +4 postinoculation with monkeypox virus) as within factor and treatment or disease outcome as between factors, respectively. Post hoc comparisons were performed using Bonferroni test. GraphPad Prism 6.01 (GraphPad Software Inc., La Jolla, CA, USA) was used for all statistical analyses.