The radioligand used in this study was the [131I]SGMIB-labelled anti-HER2 sdAb 2Rs15d, previously reported elsewhere [14]. All reagents were purchased from Sigma-Aldrich (Darmstadt, Germany) unless otherwise stated. Sodium [131I]iodide was purchased from Perkin-Elmer.
Anti-HER2 sdAb 2Rs15d was generated as described previously [16]. 2Rs15d sdAb was radiolabelled with 131I via the residualizing prosthetic group N-Succinimidyl 4-guanodinomethyl-3-[131I]iodobenzoate ([131I]SGMIB) and purified as reported previously [14].
All animal experiments were done using healthy female C57BL/6 mice (8–10-week old, 19.4 ± 1.3 g body weight mean ± standard deviation (SD)) and were conducted according to the guidelines and after approval of the Ethical Committee of the Vrije Universiteit Brussel.
2.1 Evaluation of accuracy of SPECT-based activity quantification
An animal experiment was done to evaluate the accuracy of SPECT-based activity quantification of 131I in mouse kidneys against conventional ex vivo activity measurements in a gamma counter. Mice were anesthetised by inhalation with 2% isoflurane and were intravenously injected in the tail vein with 11.5 ± 1.8 MBq [131I]-sdAb (5 µg sdAb). A total of 14 mice, divided in four groups (one group per time point), were imaged with SPECT/CT at around 1.5 (n = 3 mice), 6.6 (n = 3), 24 (n = 5) and 73 (n = 3) h post injection (p.i.) of the radioligand (Table 1). For the first two time points (time points during fast pharmacokinetics), mice were first euthanized by cervical dislocation, and SPECT-CT scans were done immediately ex vivo on their carcasses, after which both kidneys were dissected. Mice imaged at around 24 and 73 h p.i. (time points during slower pharmacokinetics) however, were imaged in vivo, euthanized and dissected immediately after SPECT-CT imaging. The kidneys of all mice were weighted, and their activity was assessed in the gamma counter. Radioactivity in kidney was corrected for the physical decay between the reference time (tref) of the activity measurement with each technique (start of GC measurement, and start and middle time of the scan for respectively ex vivo and in vivo SPECT) and the mouse time of death.
For each mouse kidney, the percentage deviation of the activity determined with SPECT (ASPECT) from the reference activity determined with gamma counting (AGC) was calculated. The mean deviation of kidneys (both left and right) of all mice per time point was calculated. Two-tailed paired t-tests were applied to the datasets of each time point to determine the significance of differences between SPECT and GC activity data. Statistical significance was defined as p < 0.05.
Additionally, for each mouse kidney the fraction of injected activity per gram of dissected tissue (FIA/g) was calculated. For each activity quantification technique, the pharmacokinetic profile was derived from the kidney FIA/g of all mice (n = 14) as a function of time p.i. (hereafter referred to as datasets GC1 for gamma counting data, and SS1 for single SPECT data), and the time-integrated activity coefficient (ã) per gram of tissue (M) was estimated (cfr. calculation details in Sect. 2.4).
Using the time-dependent fit function derived from pharmacokinetic dataset GC1, the kidney FIA/g was estimated at the time point halfway between the start and the end of in vivo SPECT scans, and at the mouse time of death, and the deviation between the resulting FIA/g values was calculated. This was used to estimate the potential bias (over-response) associated with SPECT measurements performed in vivo when compared with GC measurements, due to the estimated (expected) decrease in radioligand uptake in the kidneys during the (scan) time prior to death.
Table 1
Acquisition settings of single SPECT scans performed on 14 mice injected with the [131I]-labelled sdAb (dataset SS1), used for the comparison of kidney activity against gamma counting.
Av. time p.i. to death [max. range of variation between mice] (h) | Animal setting | No of mice, n | No of bed positions | Scan time per bed position (s) | Total effective scan time (min) | Av. (of n mice) of total scan counts acquired in photo peak window | Av. (of n mice) of photo peak counts used as background and scatter |
1.5 [± 0.0] | ex vivo | 3 | 20 | 66 | 22 | 4.5E + 06 | 5.0E + 05 |
6.6 [± 0.0] | ex vivo | 3 | 20 | 66 | 22 | 8.1E + 05 | 1.7E + 05 |
24.3* [± 0.8] | in vivo | 5 | 20 | 225 | 75 | 1.0E + 06 | 4.5E + 05 |
73.4** [± 2.0] | in vivo | 3 | 20 | 239 | 80 | 8.6E + 05 | 4.4E + 05 |
* 23.6 h average time p.i. to middle of scan. |
** 72.6 h average time p.i. to middle of scan. |
2.2 Pharmacokinetic assessment ex vivo using gamma counting
As part of the previous animal experiment, kidney uptake (in FIA/g) was assessed additionally at 3.2 and 48 h p.i. of the radioligand, but kidney activity was determined only via GC measurements. These mice (n = 3 per time point) were injected with 4.35 ± 0.09 MBq [131I]-sdAb. The time-integrated pharmacokinetic parameter ã/M was estimated (cfr. calculation details in Sect. 2.4) from all the kidney FIA/g determined with GC in this (3.2 and 48 h p.i.) and the previous section (1.5, 6.6, 24 and 73 h p.i) (hereafter referred to as dataset GC1ext), to set a reference for comparison for the ã/M values determined from longitudinal SPECT scans (cfr. section 2.3).
2.3 Pharmacokinetic assessment using longitudinal SPECT imaging
A longitudinal quantitative SPECT/CT imaging study was performed on 5 mice to evaluate the feasibility of deriving mouse-specific pharmacokinetic information of the [131I]-sdAbs. All mice were intravenously injected with 13.62 ± 1.15 MBq [131I]-sdAb. Sequential SPECT-CT scans were performed in vivo on each mouse, starting at approximately 1, 3, 6, 24, 46 and 70 h p.i. of the radioligand (cfr. acquisition settings in Table 2). Immediately after the last scan, mice were euthanized and both kidneys were dissected and weighted. Kidney activities were determined from SPECT images (cfr. section 2.5) using as a reference time the middle time of each SPECT scan. The FIA/g was calculated at each scan time point, and the time-integrated pharmacokinetic parameter ã/M was estimated (cfr. calculation details in Sect. 2.4) from the pharmacokinetic profile of each mouse (hereafter referred to as datasets LS1, LS2, LS3, LS4 and LS5, i.e. one dataset per mouse).
Table 2
Acquisition settings of the longitudinal SPECT scans performed in vivo on 5 mice injected with the [131I]-labelled sdAb, corresponding to datasets LS1-LS5.
Av. scan start time p.i. [max. range of variation between mice] (h) | No of bed positions | Scan time per bed position (s) | Total effective scan time (min) | Av. (of n mice) of total scan counts acquired in photo peak window | Av. (of n mice) of photo peak counts used as background and scatter |
1.0 [± 0.0] | 20 | 51 | 17 | 4.9E + 06 | 5.2E + 06 |
3.3 [± 0.0] | 20 | 66 | 22 | 1.7E + 06 | 2.4E + 06 |
6.0 [± 0.1] | 20 | 66 | 22 | 1.1E + 06 | 2.0E + 05 |
24.2 [± 2.0] | 20 | 150 | 50 | 8.2E + 05 | 3.0E + 05 |
46.3 [± 3.7] | 20 | 239 (159)* | 80 (53)* | 9.1E + 05 (6.0E + 05)* | 4.4E + 05 (3.1E + 05)* |
70.2 [± 3.3] | 20 | 239 (159)* | 80 (53)* | 7.8E + 05 (5.2E + 05)* | 4.0E + 05 (2.8E + 05)* |
* specification used for the two mice from datasets LS1 and LS2. |
2.4 Pharmacokinetic modelling
Kidney pharmacokinetic data were analysed by nonlinear least squares fitting (MATLAB R2019a, MathWorks, Massachusetts, USA) to a mathematical function of time elapsed p.i. (t). A negative power function with two coefficients (c1 and c2) (Eq. 1) was chosen among the various mathematical functions considered (cfr. supplementary data). Eight datasets of FIA/g as a function of time (datasets GC1, SS1, GC1ext and LS1-LS5) were analysed. The Pearson’s correlation coefficient (R2) was used to quantify goodness of fit.
The time-integrated pharmacokinetic parameter ã/M was calculated for the left kidneys from mathematical integration of the fit function from 1.5 h to 72 h only, since the purpose of estimating ã/M was only to compare different datasets and this was the common period p.i. covered by most datasets.
2.5 SPECT-CT imaging and image quantification
SPECT-CT acquisitions were performed with a VECTor/CT small-animal SPECT-CT system (U-PET/SPECT4, CT Skyscan1178; MILabs, Utrecht, Netherlands) [17]. For photon detection the SPECT module uses three stationary large NaI(Tl) detectors each with a thickness of 18 mm. A high-energy mouse collimator with 114 pinholes of 1.6-mm diameter was used for imaging of 365-keV photons of 131I. A scan field-of-view with a 20-mm axial length located on the abdominal region comprising the kidneys was used for all mouse SPECT scans. Additional SPECT acquisition settings are reported in Tables 1 and 2. Following a SPECT scan, a whole-body CT scan (55 kV X-ray tube voltage, 615 µA tube current) was acquired.
SPECT and CT images were generated and co-registered with VECTor’s manufacturer-provided software. SPECT images with 0.6-mm-wide cubic voxels were reconstructed using a 20% photo-peak window centred at 365 keV, using VECTor’s pixel-based ordered-subset expectation maximization (OSEM) iterative reconstruction algorithm [18]. Thirty iterations with two subsets (i.e. 60 OSEM updates, equivalent to 60 maximum-likelihood expectation-maximization iterations) were performed for all scans. A prior investigation indicated that 60 OSEM updates were sufficient to ensure convergence in the recovery of hot rods with a diameter in the range of 4 to 6 mm, which we consider to be comparable to the size of mouse kidneys. A system response matrix optimized for 364 keV photons was used [19]. Two adjacent background-and-scatter windows of approximately 6% width were used for Compton photon scatter and background signal correction using the triple-energy-window method [9]. Additionally, SPECT images were registered to CT images and the resulting SPECT images were corrected for photon attenuation based on CT data [10]. No post-reconstruction filters (for smoothing) were used. The final resampled SPECT images used for analysis have nearly cubic voxels of approximately 0.17-mm width.
SPECT voxel data (count rate, in reconstructed “cps”) were calibrated in terms of activity concentration (MBq.mL− 1) using a calibration factor determined from the SPECT scan of a syringe containing a [131I]-NaI solution with a calibrated activity concentration directly traceable to high-resolution gamma spectrometry (more details in the supplementary data).
Image quantification was done with AMIDE 1.0.4 [20]. The amount of 131I activity in the kidneys was determined from activity-calibrated SPECT images using ellipsoidal VOI. A VOI was manually drawn over each kidney based on the CT image, and its position and proportions were then fine-tuned visually according to the SPECT image. The VOI volume was adjusted to match the kidney volume estimated from the mass of the dissected kidney and an assumed tissue density of 1.04 g.mL− 1.
2.6 Gamma counting, GC
The activity of dissected tissues was measured in a Cobra II model 5003 gamma counter (Canberra-Packard, Schwadorf, Austria) using a measurement protocol optimized to limit the overall measurement uncertainty (cfr. details in the supplementary data).
The activity calibration of GC measurements was directly traceable to high-resolution gamma spectrometry (cfr. section 2.7). The calibration procedure of GC (cfr. details in the supplementary data) was reproducible within a 2.1% relative standard deviation, corresponding to an expanded uncertainty of ± 3.4% expressed at the 95.5% confidence interval (CI) (k = 3.31 for a t-distribution with 3 degrees of freedom (4 stock solutions)).
The combined expanded uncertainty of gamma counting tissue activity measurements (UGC) was always within ± 3.6% (expressed at the 95.5% CI), and was determined from the square root of the summation in quadrature of the uncertainty due to calibration reproducibility and the counting statistical error of the tissue sample.
2.7 Reference measurements for SPECT and GC activity calibrations
The reference activity concentrations of all 131I stock solutions used for SPECT and GC activity calibrations were determined by high-resolution gamma spectrometry analysis using a high-purity germanium detector (model GC1818-7500SL; Mirion-Canberra, Meriden, USA) calibrated for photon energy and detection efficiency. More details on the measurement procedure are provided in the supplementary data. The relative statistical uncertainty of the reference activity concentration of each of the calibration stock solutions was always within 1.6% at 95.5% CI (coverage factor k = 2).
2.8 Determination of SPECT activity recovery coefficients
Activity recovery coefficients (ARC) were calculated based on a SPECT study of a phantom filled with a [131I]-NaI solution with a calibrated activity concentration of 10.4 MBq.mL− 1. The phantom is a 26-mm diameter cylindrical container made of acrylic plastic, with two fillable compartments containing various fillable rods. Both compartments (~ 4.4 mL) were filled, but only one compartment, containing two rods of 4- and 6-mm diameter and 32-mm length, was used for analysis. A 20-min SPECT scan of the whole phantom (120 bed positions, 10 seconds per position) was acquired. SPECT images were generated with the same image reconstruction and correction settings as for mouse scans. A cylindrical VOI was defined at the centre of each rod with a diameter equal to the diameter of the rod and a length of 4 mm. ARC were calculated, for each rod, as the ratio of the mean activity concentration of the VOI and the reference activity concentration of the 131I solution at the start of the SPECT scan.
To assess the impact of lower counts on activity recovery, the phantom scan was reconstructed using only 10.0%, 5.0%, 2.5%, 1.0%, 0.5% and 0.1% of the counts from the list-mode data. These reconstructions emulate scans with lower activity concentrations equivalent respectively to 1.044, 0.522, 0.261, 0.104, 0.052 and 0.010 MBq.mL− 1.
Additional cylindrical VOIs, each with 4-mm height and 4-mm diameter, were drawn along the 4-mm rod. These VOIs were repeated on 6 planes (i.e. 7 VOIs in total) for an axial distance of 30 mm. For each activity concentration reconstructed, the statistical uncertainty of activity recovery (URec) due to the regional (VOI) variability of the SPECT image counts within the rods was estimated, at the 95.5% CI, as:
Where σVOIs and ĀVOIs are respectively the standard deviation and the mean of the VOI activities determined in all the VOIs (nVOIs=7 VOIs); and k is the coverage factor, which is equal to 2.52 for a t-distribution with 6 (nVOIs-1) degrees of freedom. This metric captures the statistical uncertainty in the activity recovery in the VOIs, and is used in this study as an indicator of the precision of the SPECT reconstruction to recover the activity in mouse kidneys with varying levels of activity concentration.