Introduction
O-GlcNAcylation is a common post-transcriptional glycosylation that occurs extensively at the intracellular level of the brain [
1]. This reversible cycling is modulated by two enzymes;
O-linked β-
N-acetylglucosamine (
O-GlcNAc) transferase attaches
O-GlcNAc to a protein, whereas
O-GlcNAcase (OGA) removes it [
2]. Since the discovery of
O-GlcNAcylation of tau and its impact on tau phosphorylation,
O-GlcNAcylation-related studies have rapidly increased in Alzheimer’s disease (AD) [
3]. Notably, because hyperphosphorylated tau forms aggregates and neurofibrillary tangles—one of the hallmarks of AD [
4]—
O-GlcNAcylation and its relationship to phosphorylation have become the subjects of considerable investigative interest with regard to a group of neurodegenerative diseases collectively called tauopathies.
O-GlcNAcylation has been known to stabilize microtubule-associated protein tau by hampering hyperphosphorylation and aggregation [
5]. Indeed, initial ex vivo research supported the reciprocal relationship
O-GlcNAcylation and phosphorylation [
6], and analysis of human brain tissue revealed lower levels of tau-specific and overall
O-GlcNAc in individuals with AD [
7]. Furthermore, a series of animal studies found that increased
O-GlcNAcylation cycling by an OGA inhibitor reduced tau protein aggregation and restored cognitive function [
8‐
10]. Not surprisingly, rising levels of brain
O-GlcNAcylation has been explored as a potential therapeutic strategy for attenuating the progression of AD and other tauopathies; as a result, potent and selective OGA inhibitors have been developed, and early phase clinical trials with these agents have begun [
11,
12].
In this context, a positron emission tomography (PET) radioligand capable of imaging OGA could improve our understanding of the pathophysiology of neurodegenerative diseases, provide evidence of drug-target engagement, and help with dose selection of therapeutic candidates. Previous studies from our laboratory reported initial PET results of [
18F]LSN3316612 (
N-(5-(((2
S,4
S)-2-methyl-4-(6-fluoropyridin-2-yloxy) piperidin-1-yl)methyl)thiazol-2-yl)acetamide) in both nonhuman primates and healthy human volunteers. In nonhuman primates, [
18F]LSN3316612 exhibited excellent properties for quantifying OGA in the brain, including high brain uptake with a high proportion of specific binding, lack of radiometabolite interference with quantification, and feasibility of quantification with compartmental modeling [
13]. A follow-up study reported similarly promising results in the brains of eight healthy volunteers [
14]; human brain uptake was well quantified with compartmental modeling and showed no evidence of accumulation of radiometabolites.
This study sought to further evaluate the suitability of [18F]LSN3316612 for use in human clinical research. Brain scans were evaluated from an additional set of volunteers (for a combined sample of 17). In addition, test-retest imaging was conducted in 10 volunteers, and 6 whole-body scans were conducted to measure radiation exposure to organs of the body.
Material and methods
Radioligand synthesis
[
18F]LSN3316612 was synthesized by a nucleophilic substitution reaction on a nitro-aryl precursor, as previously described [
13]. The nitro precursor (
N-(5-(((2
S,4
S)-2-methyl-4-(6-nitropyridin-2-yloxy)piperidin-1-yl)methyl)thiazol-2-yl)acetamide) was provided at Eli Lilly.
Participants
The entire population for both brain and whole-body imaging consisted of 11 male and 12 female healthy volunteers; 17 of these had brain scans and six had whole-body scans (Table
1). All volunteers were free of current medical or psychiatric illnesses as determined by medical history, physical examination, electrocardiogram, urinalysis, and laboratory blood tests (complete blood count, serum chemistries, and thyroid function test). The volunteer’s vital signs were recorded before radioligand injection and at 15, 30, 90, and 120 min post-injection.
Table 1Demographic characteristics and PET scan parameters for 17 healthy volunteers injected with [18F]LSN3316612
Male:female (n) | 8:9 | 5:5 | 5:5 | 3:3 |
Age (years) | 40 ± 11 | 43 ± 11 | 43 ± 11 | 34 ± 16 |
Body weight (kg) | 73 ± 17 | 72 ± 16 | 75 ± 17 | 69 ± 12 |
Injected activity (MBq) | 187 ± 6 | 185 ± 7 | 189 ± 3 | 172 ± 47 |
Molar activity (MBq/nmol) | 55 ± 22 | 45 ± 16 | 50 ± 13 | 64 ± 9 |
Injected mass dose (nmol/kg) | 0.058 ± 0.033 | 0.069 ± 0.038 | 0.057 ± 0.016 | 0.040 ± 0.011 |
Brain image acquisition and processing
Brain PET scans were acquired from 17 healthy volunteers with an mCT scanner (Siemens Medical Solution, Cary, NC, USA). After a low-dose CT scan for attenuation correction, [18F]LSN3316612 (188 ± 5 MBq) was intravenously injected and PET data were acquired for 180 min with concurrent arterial blood sampling. Data were reconstructed into 45 frames (6 × 0.5 min, 3 × 1 min, 2 × 2 min, 34 × 5 min) using a three-dimensional ordered subset expectation-maximization algorithm. Brain uptake was expressed as a standardized uptake value (SUV), which normalizes for injected radioactivity and body weight. For structural magnetic resonance (MR) imaging, all participants underwent sagittal T1-weighted brain MR, using a 3T Philips Achieva scanner (Bothell, WA, USA) with turbo field echo sequence (repetition time = 8.1 ms, echo time = 3.7 ms, flip angle = 8, matrix = 181 × 256 × 256, voxel size = 1 × 0.983 × 0.983 mm).
Image pre-processing—such as coregistration between PET and MR, segmentation, and atlas normalization—was performed with the PNEURO pipeline implemented in PMOD 3.903 (Zurich, Switzerland). A total of 83 volumes of interest were defined based on the Hammers’ probabilistic brain atlas [
15] and the subject’s individual MR image and subsequently combined into an individual template consisting of 16 regions that encompass the entire lobes of the brain and the principal subcortical structures: frontal, parietal, temporal, occipital, insula, amygdala, hippocampus, cingulate, striatum, thalamus, globus pallidus, brainstem, corpus callosum, cerebellar cortex, cerebellar white matter, and cerebral white matter. Regional time-activity curves were obtained by applying the template on the dynamic PET images transformed into MR space. The quality of PET-MR coregistration was visually assessed by a side-by-side comparison of PET, MR, and fused images at the end of PNEURO pipeline.
For the test-retest study, 10 out of the 17 volunteers who had a brain scan were scanned again on a different day under identical procedures. The interval between test and retest scans ranged from 8 to 150 days.
Measuring [18F]LSN3316612 in plasma
During the brain PET scan, arterial blood was continuously monitored for 10 min at a rate of 5 mL/min, and radioactivity was measured with a cross-calibrated coincidence detector (PBS-101, Comecer, The Netherlands). Manual arterial samples were also obtained at 3, 5, 10, 15, 30, 60, 90, 120, 150, and 180 min after [18F]LSN3316612 injection.
For all blood samples, plasma concentrations of [
18F]LSN3316612 were measured using an automatic gamma counter and were corrected after separation from radiometabolites using high-performance liquid chromatography (HPLC), as previously described [
16], but with an X-Terra C18 column (10 μm, 7.8 × 300 mm; Waters Corp., Milford, MA, USA) and a mobile phase of MeOH:10 mM ammonium formate (75% by volume). The extraction efficiency of the deproteinization method (e.g., acetonitrile extraction) was quantified using the radioactivity of the precipitate. The mean extraction efficiency of the 10 blood samples obtained from each of the 10 participants during a 180-min test scan was 79.6% ± 6.4% (
N = 10) which showed no significant difference from the 82.5% ± 5.7% mean that was obtained during the retest scans (
P = 0.101). The plasma free fraction (
fP; the non-protein bound fraction) was measured by ultrafiltration [
16]. Using the blood-to-plasma ratios determined from the manual samples, total plasma radioactivity curves were obtained from measured whole blood data from automatic sampling for the first 10 min, then scaled to fit manual sample data. Total plasma radioactivity and whole blood activity were then fitted to a tri-exponential function. A Hill function [
17] was used to fit the unchanged parent fraction. A time-activity curve of radiometabolite-corrected plasma parent radioactivity was generated implicitly by the product of the total plasma activity curve and parent fraction in PMOD, which was used as the input function.
Tracer kinetic modeling
All kinetic analyses, including fitting blood curves, were conducted with the PKIN module in PMOD. The outcome measure, total distribution volume (VT), was calculated using one-and two-tissue compartmental models with noise equivalent count weighting and the radiometabolite-corrected arterial input function fitted to a tri-exponential function. Whole blood curves were used to correct for activity in the vascular component, assuming that blood volume was 5% of total brain volume. An optimal kinetic model was determined based on the relative fitness of the model (i.e., Akaike information criterion (AIC) and F test) and the identifiability of VT (i.e., percent standard error (%SE) estimated from the theoretical parameter covariance matrix).
Test-retest variability and reliability
Test-retest variability and absolute test-retest variability between test and retest scans were calculated as follows:
$$ \mathrm{Test}-\mathrm{retest}\ \mathrm{variability}\ \left(\%\right)=\frac{\mathrm{Test}\ \mathrm{value}-\mathrm{Retest}\ \mathrm{value}}{\left(\mathrm{Test}\ \mathrm{value}+\mathrm{Retest}\ \mathrm{value}\right)/2}\times 100 $$
$$ \mathrm{Absolute}\ \mathrm{test}-\mathrm{retest}\ \mathrm{variability}\ \left(\%\right)=\frac{\mid \mathrm{Test}\ \mathrm{value}-\mathrm{Retest}\ \mathrm{value}\mid }{\left(\mathrm{Test}\ \mathrm{value}+\mathrm{Retest}\ \mathrm{value}\right)/2}\times 100 $$
To assess test-retest reliability, the intra-class correlation coefficient (ICC) of each region was calculated as follows [
18]:
$$ \mathrm{ICC}=\frac{\mathrm{BSMSS}-\mathrm{WSMSS}}{\mathrm{BSMSS}+\mathrm{WSMSS}} $$
where BSMSS and WSMSS are the mean sums of squares between subjects and within subjects, respectively. In the test-retest study, ICC value could range from – 1 to 1, and values closer to 1 indicated better reliability [
19].
Whole-body biodistribution and radiation dosimetry
To determine the radiation exposure to body organs, a separate group of six healthy volunteers underwent a whole-body PET scan after intravenous administration of [18F]LSN3316612 (172 ± 47 MBq). Dynamic scans were acquired using the same mCT scanner in seven contiguous segments from top of the head to mid-thigh in 14 frames of increasing duration (75 s to 15 min) for a total scan time of 120 min.
Thirteen source organs that could be identified as hot uptake foci on PET images were generously delineated on the tomographic images to ensure that all accumulated radioactivity in each organ was encompassed using PMOD: brain, heart, lungs, spleen, liver, kidneys, gallbladder, red marrow, stomach, testes/ovaries, urinary bladder, and small intestine. Uptake in the source organs was corrected with a recovery coefficient based on the average activity of the frames of the whole-body dynamic scan using large regions of interest drawn semi-automatically around the body. The average recovery coefficient of the 6 volunteers was 90%. The radioactivity concentration, measured without decay correction, was expressed as a percentage of injected dose for each organ. The organ residence time was calculated as the area under the time-activity curve using the trapezoid rule and physical decay after the last frame. Because the total red marrow present in the lumbar vertebrae accounts for ~ 12.3% of the mass of red marrow in the whole body [
20], the red marrow final residence time was obtained by dividing the residence time in the lumbar vertebrae by 0.123. Values for the gastrointestinal tract were generated by the International Commission for Radiation Protection (ICRP) model in OLINDA/EXM1.1 [
21] as activity entering the small intestine [
22]. To obtain the residence time for the remainder of the body, residence times for all source organs were summed and then subtracted from the theoretical value of 2.65 h (=
18F half-life/ln2).
Statistical analysis
Quantitative results are presented as mean ± standard deviation (SD) unless otherwise noted. Differences between the test and retest groups were analyzed using a two-way t test, and those among the three groups of volunteers using the one-way analysis of variance (ANOVA). Bonferroni correction was used for multiple comparisons of each between-group comparison. The correlation between continuous variables was evaluated with linear regression analysis. Statistical significance was set at P < 0.05, and all statistical analyses were conducted using GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA).
Discussion
The present study found that [18F]LSN3316612 was an excellent PET radioligand for quantifying OGA in the human brain with the exception of an unexplained increase in VT on retest scanning in nine of 10 volunteers. Brain uptake was generally high and could be quantified as VT with excellent identifiability using a two-tissue compartment model. [18F]LSN3316612 exhibited good absolute test-retest variability (~ 12.5%), but the arithmetic test-retest variability was far from 0 (~ 11.3%), reflecting an almost uniform increase of VT on retest scanning. VT values were stable after 110 min of scanning in all regions, suggesting that radiometabolites did not accumulate in the brain. Efforts to investigate an alternative quantification method for [18F]LSN3316612 binding without blood to assess the possibility of eliminating arterial sampling found that measurements obtained using only brain activity (i.e., AUC from 150 to 180 min) were strongly correlated with regional VT values within an individual volunteer during test and retest conditions (R2 = 0.84), and similar reliability to VT was observed (ICC = 0.68 for AUC150–180 and 0.64 for VT).
Despite the promising characteristics of [
18F]LSN3316612, the unexplained increase of
VT under retest conditions requires further investigation before the radioligand can be widely used. In this study, the “gold standard” measurement of OGA density (i.e.,
VT) in retest scans was greater than that in the test scans in 9 out of 10 volunteers, with an average increase of 12.1%. This unexpected increase may reflect two possibilities. The first is that OGA levels actually increased on retest scans, and the second is that our measurement of
VT was flawed. With regard to the first possibility, several factors were examined that might correlate with and thereby explain an increase in retest scan. These included interval between test and retest, seasonal or diurnal variations, molar activity, injected mass dose of the test scan,
fP, image acquisition and processing, and the volunteers’ laboratory test results (Additional file
2: Table S2). During the entire course of the study, factors such as variation of dose calibrator, well counter, and HPLC were all closely monitored, and we found no significant change in the external detectors related to radioactivity measurement. Therefore, none of these factors, we believe, contributed to the increase of
VT. Nevertheless, the sample size was small (
n = 10) and may have had inadequate power to detect true effects.
The second, and more likely, the possibility is that our measurement of VT was flawed in some way. Indeed, the data suggested potential errors in measuring the input function, i.e., the concentration of parent radioligand in plasma over time. In particular, the variability and reliability of only brain activity (i.e., AUC150–180) was as good as or slightly better than that of VT, which uses brain and plasma data. Addition of mean parent concentration in plasma during the last 30 min (i.e., AUC/CP150–180) worsened the variability and reliability in comparison to only AUC150–180. To explore whether measurement of the input function was flawed, we measured the AUC from time zero to infinity (AUC0–∞) of the parent radioligand in plasma to determine how it may have affected VT. Please note that VT = AUC0–∞ of brain curve/AUC0–∞ of the plasma curve. We found that the AUC0–∞ of the plasma curve was 8.6% lower on retest compared to test scans, which would explain most of the increase in VT (12.1%). To determine which component of the plasma curve might have been measured inaccurately, we explored tri-exponential fitting of the plasma curve, which identified three half-lives (1.0 ± 0.4, 8.0 ± 3.3, and 140.7 ± 39.2 min, respectively). The third (slowest) component contributed most (76.3%) of the AUC0–∞. Thus, sampling for only 180 min might inadequately define an exponent with a half-life of 141 min. In addition, the concentration of radioactivity in plasma was decreasing and difficult to measure. As a result, we suspect, but are not certain, that errors may have occurred in the measurement of the plasma input function to increase VT on retest scans. However, we cannot explain why a consistent bias would have existed so that plasma AUC0–∞would be lower on retest. A study with a larger sample size and longer plasma sampling might help answer these questions.
The average VT values in the 17 volunteers who had brain scans reached 90% of terminal 3-h values at 110 min and remained stable thereafter. This indicates that no troublesome radiometabolites entered the brain. Regions with high VT (e.g., hippocampus) were slower to reach stable VT than regions with lower VT (e.g., cerebellum), consistent with the notion that higher enzyme density regions require a longer time to reach equilibrium. The evaluation of time stability revealed that the [18F]LSN3316612 PET scan time could be reliably shortened to 120 min; however, time stability would need to be re-evaluated within disease-specific groups because the equilibrium time might change depending on the degree of reduced OGA expression and decreased blood flow that can occur in tauopathies.
Following the injection of [
18F]LSN3316612, the whole-body distribution of radioactivity reflected both the distribution of OGA and the metabolism of the radioligand. OGA is expressed in relatively high density in the brain, gastrointestinal tract, exocrine glands, and immune system [
24,
25]. Accordingly, whole-body [
18F]LSN3316612 PET exhibited high accumulation in the brain, small bowel, salivary glands, and spleen. Similarly, uptake in the lumbar vertebrae, which contain approximately 12.3% of all red marrow in the body [
20], likely reflected uptake in hematopoietic stem cells in the red marrow [
25]. Most of the radioactivity was in the bone marrow and not in distal appendicular bones without red marrow, suggesting that negligible defluorination, if any, occurred. In addition to reflecting the distribution of OGA in the body, radioactivity also likely reflected metabolism due to the hepatobiliary tract excretion of radiometabolites via urine—e.g., a glucuronidated radiometabolite of the lipophilic radioligand. The urinary bladder also received relatively high exposure, although this was probably due to long retention of urine during the 3-h PET scan and could therefore be modestly lowered during a typical 2-h scan and by immediate voiding after completion of PET scan. The effective dose (μSv/MBq) of [
18F]LSN3316612 (20.5 ± 2.1) was similar to that of other
18F-radioligands used for brain imaging (20.8 ± 6.7 for 21 radioligands) [
23].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.