Non-invasive modelling and parametric methods for quantification of MAO-B activity using [ 11 C]L-deprenyl-D2 PET

Abstract

The PET radioligand [¹¹C]L-deprenyl-D2 binds to monoamine oxidase B (MAO-B), enabling studies of neuroinflammation in various neurological disorders. This study aimed to evaluate non-invasive and parametric methods for quantifying [¹¹C]L-deprenyl-D2 PET. Twenty-eight subjects underwent 60 min dynamic [¹¹C]L-deprenyl-D2 PET. The irreversible binding parameter λk₃ and the net influx rate relative to the non-displaceable volume of distribution K_ND were estimated with an irreversible two-tissue plasma-input compartment model (2T3k). Non-invasive quantification of K_ND was performed using cerebellum corrected for specific binding with previously suggested corrections as reference tissue. These corrections were further optimised by comparison to the 2T3k non-displaceable signal in cerebellum. Parametric images were computed using a basis function implementation of the 2T3k model, reference Patlak methods and the reduced reference tissue model. Additionally, a semi-quantitative specific binding index (SBI) was calculated as the ratio of the mean activity concentration at 30–40 or 50–60 min p.i. in target tissues to the peak cerebellum uptake. Parametric λk₃ images agreed well with VOI-based nonlinear regression. Modified cerebellar reference models and SBI demonstrated poor-to-moderate correlations with arterial input-based analysis and yielded false positive differences in voxel-wise group comparisons. Full kinetic analysis with arterial input is required for accurate assessment of MAO-B activity with [¹¹C]L-deprenyl-D2.

Keywords

kinetic modelling MAO-B neuroinflammation parametric images reference tissue models

Introduction

The radioligand [¹¹C]L-deprenyl-D2 is used in positron emission tomography (PET) to study neuroinflammation. It binds irreversibly to the enzyme monoamine oxidase B (MAO-B), which is primarily located in astrocytes.^1–4 L-deprenyl (selegiline) is a selective and irreversible inhibitor of MAO-B.^5,6 Elevated levels of MAO-B density and activity serve as biological markers for various neurological disorders and brain abnormalities. Previous studies have demonstrated the utility of labelled L-deprenyl with PET for mapping of MAO-B in the human brain,^7–9 and clinical PET studies have shown its application in Alzheimer’s disease,^10–14 epilepsy,^15–17 Creutzfeld–Jakob disease,¹⁸ amyotrophic lateral sclerosis,¹ traumatic brain injury¹⁹ and evaluation of MAO-B inhibitors.^20–23 This technique also holds relevance for studying multiple sclerosis (MS), a chronic inflammatory disease of the central nervous system characterised by demyelination of neuronal axons. It manifests in two forms: relapsing-remitting MS, which may eventually progress into secondary progressive MS (SPMS).^2,24–26 In the context of MS, in vivo imaging and quantification using PET serve as a valuable tool for the diagnostic detection of reactive astrogliosis.

The gold standard method for quantifying [¹¹C]L-deprenyl-D2 is the irreversible two-tissue compartment model (2T3k), which requires a metabolite-corrected plasma input function derived from arterial sampling.^8,20,27 Although the irreversible trapping rate constant k₃ can be considered as the primary measure of MAO-B activity, Fowler et al.⁸ proposed the macroparameter λk₃, defined as λk₃ = (K₁/k₂) k₃, as a more robust alternative to k3 alone.^28,29 There are several reasons for favouring λk₃ over k₃ or the more commonly used net influx rate (K_i = K₁k₃/(k₂ + k₃)). First, bias can occur in estimating k₃ and K_i because estimates of k₂ and k₃ are correlated, especially in regions of high MAO-B concentrations where k₂ may be underestimated, leading to underestimation of k₃. The issue of correlation between k₂ and k₃ is mitigated by including their ratio k₃/k₂. Secondly, due the high extraction of L-deprenyl, K₁ is influenced by blood flow, reducing sensitivity of K_i to changes in MAO-B levels. In contrast, the macroparameter λk₃ is, unlike K_i, independent of blood flow as it incorporates the ratio K₁/k₂.^8,28–30

Non-invasive [¹¹C]L-deprenyl-D2 quantification would significantly enhance its utility by eliminating the need for arterial blood sampling and metabolite analysis, which can be both cumbersome and prone to errors. However, the widespread presence of MAO-B throughout the brain means that there is no region entirely devoid of MAO-B, and hence no suitable reference region. Therefore, two methods have previously been proposed to correct for cerebellar MAO-B binding when using a reference tissue model. One approach involves multiplying the cerebellar time-activity curve (TAC) by an exponential function,¹⁶ while the other utilises convolution subtraction of an exponential function.¹ The latter method in particular has been extensively used in studies of neurodegenerative diseases involving [¹¹C]L-deprenyl-D2,^{11–14,31–33} but a validation of this method by comparison to plasma-input data has never been published.

Voxel-wise quantification of [¹¹C]L-deprenyl-D2 enables the generation of parametric images of MAO-B binding for visualising disease pathology. However, to the best of our knowledge, neither non-invasive methods nor voxel-wise quantification have been thoroughly validated for [¹¹C]L-deprenyl-D2. Therefore, the present study aimed to evaluate and optimise non-invasive and semi-quantitative methods for kinetic analysis and to validate parametric methods for [¹¹C]L-deprenyl-D2 PET. First, we will compare the previously suggested non-invasive methods and proposed improvements to these methods, including a semi-quantitative uptake measure, to plasma-input analysis. Then, we will validate parametric imaging methods by comparison to volume of interest (VOI) based methods. Finally, the limitations of non-invasive methods will be addressed by simulations assessing the effect of violations of the underlying assumptions of reference tissue models.

Material and methods

Subjects

Data was used from 28 subjects included in a study on MS and neuroinflammation: nine patients with SPMS, eight individuals who had undergone autologous haematopoietic stem cell transplantation (AHSCT) as a treatment for MS,³⁴ and 11 healthy controls. SPMS and AHSCT patients were recruited from the Department of Neurology at Uppsala University Hospital, Sweden, while healthy controls were recruited through local and social media advertisements. The age and sex distribution was similar across the patient groups and healthy controls (mean age: 39 ± 10 years, range: 19–56 years; 19 females). The results of the evaluation of MAO-B activity within and between these groups will be presented in a separate publication. All subjects provided written informed consent prior to inclusion, and the study was conducted in accordance with the Declaration of Helsinki and approved by the Regional Board of Medical Ethics in Uppsala (EPN 2014/453/1).

Image acquisition

Each subject underwent a 60 min dynamic PET scan following a controlled intravenous bolus injection of ~6 MBq/kg (mean 407 ± 57 MBq) [¹¹C]L-deprenyl-D2^30,35,36 (5 ml tracer solution at 1 ml/s followed by 35 ml saline at 2 ml/s). The subjects fasted for 4 h and were instructed to abstain from coffee, alcohol and tobacco for 12 h prior to the examination. The scans were performed on either an ECAT Exact HR+ (Siemens/CTI, Knoxville, TN; n = 12) stand-alone PET scanner with a 15.5 cm axial field of view (FOV)³⁷ or a Discovery MI digital 4-ring PET/computed tomography (CT) scanner (GE Healthcare, Waukesha, WI; n = 16) with a 20 cm axial FOV.³⁸ The dynamic PET data was corrected for attenuation either with a ⁶⁸Ge-transmission scan (ECAT Exact HR+) or with a low-dose CT scan (Discovery MI). Images were reconstructed into 22 frames (6 × 10, 3 × 20, 2 × 30, 2 × 60, 2 × 150, 4 × 300, 3 × 600 s). Reconstruction parameters were selected to result in matching image spatial resolution of ~6.5 mm between the two scanners: ordered subsets expectation maximisation (OSEM) with six iterations, eight subsets and a 4 mm Hanning filter for the ECAT Exact HR+, and three iterations, 16 subsets and a 5 mm Hanning filter for the Discovery MI.

In addition, three-dimensional (3D) T1-weighted images were acquired on an Achieva dStream 3T magnetic resonance imaging (MRI) system (Philips Healthcare, Best, The Netherlands), using a 32-channel head coil. The T1-weighted gradient echo sequence had an acquisition duration of 3 min 50 s, a repetition time of 8.3 ms and an echo time of 3.8 ms, and a spatial resolution of 0.9 × 0.9 × 1.0 mm³ with 220 slices.

Arterial blood sampling and metabolite analysis

Continuous on-line arterial blood sampling was performed during the initial 10 min of the scan (3 ml/min) using an ABSS V3 (Allogg, Mariefred, Sweden; ECAT Exact HR+) or PBS-100 (Veenstra, Joure, The Netherlands; Discovery MI). Discrete blood samples (3 ml) were drawn for metabolite analysis at 5, 10, 20, 30, 40 min p.i., and for the measurement of whole-blood and plasma radioactivity (2 ml) at 5, 10, 15, 20, 25, 30, 35, 40, 50 and 60 min p.i. Discrete blood samples were measured in a well counter that was cross-calibrated with the PET scanners. The continuous arterial sampling data were calibrated using the discrete samples taken at 5 and 10 min p.i. and extended to 60 min by a multi-exponential fit to the discrete data. The input function was then derived by multiplying the measured whole-blood data by a single exponential fit to the plasma-to-whole-blood ratios and a sigmoid function fit to the measured fraction of unmetabolised [¹¹C]L-deprenyl-D2 in plasma. Finally, the input function was corrected for delay relative to the whole-brain TAC.

Volumes of interest

Dynamic PET scans were corrected for inter-frame motion using in-house developed software in MATLAB (The MathWorks, Inc., Natick, MA). The 3D T1-weighted MR images were co-registered to the PET images using rigid transformation and segmented into grey matter, white matter and cerebrospinal fluid using SPM8 (Statistical Parametric Mapping; Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, London, UK). An automatic probabilistic VOI template implemented in PVElab software³⁹ was used to define grey matter VOIs. The data analysis included the striatum, thalamus, cerebellum and a global cortical region comprising the frontal, temporal, parietal, occipital, anterior cingulate and posterior cingulate cortex. The VOIs were overlaid onto the dynamic PET images to extract TACs.

Tracer kinetic analysis

The irreversible two-tissue compartment model (2T3k) with plasma input function and a blood volume parameter was fitted to all TACs using weighted nonlinear regression.^27,30,40 Parameter estimates of the model rate constants K₁, k₂ and k₃ were obtained along with the macroparameter λk₃,⁸ the non-displaceable volume of distribution V_ND (= K₁/k₂), which equals λ, and the net uptake rate relative to the non-displaceable volume of distribution K_ND (= k₂k₃/(k₂ + k₃)). This last parameter has also been referred to as K, K_i or K_iref in other studies.

Non-invasive estimation of K_ND was performed using Patlak analysis^41,42 over a time interval of 10–60 min and the reduced reference tissue model (RRTM),⁴³ using reference tissue TACs as described in the next section. To further evaluate the validity of a reference tissue approach with a corrected cerebellum TAC, Spearman’s correlation, mean and standard deviation of relative differences were calculated between V_ND values in the striatum and cerebellum.

Additionally, a specific binding index (SBI) was calculated as the ratio of the mean activity concentration at 30–40 or 50–60 min p.i. in target tissues to the peak value of the cerebellum TAC, providing a simplified non-invasive semi-quantitative approach without the need for kinetic modelling.

Non-invasive methods were validated by comparison to λk₃ values derived from the 2T3k plasma-input model using Spearman’s correlation. In addition, non-invasive K_ND values were compared to 2T3k K_ND using Spearman’s correlation and Deming regression.

Reference time–activity curve

A theoretical non-displaceable reference curve, reflecting only the non-displaceable fraction of the cerebellar TAC, was calculated for each subject using the single-tissue compartment model (1T2k) and the estimated K₁ and k₂ from the 2T3k plasma-input model. Four different methods were used to correct the cerebellar TAC for MAO-B binding, of which two have been previously published (A and B):

Multiplication of the cerebellar TAC with an exponential function, using the following equation, as previously proposed by Bergström et al.¹⁶:

A : C_{ref} (t) = C_{cer} (t) \cdot e^{- 0.04 t}

(1)

where C_cer is the cerebellar grey matter radioactivity concentration measured at time t and C_ref is the cerebellar grey matter radioactivity concentration after correction for MAO-B binding.

A modification of equation (1):

A' : C_{ref} (t) = {\begin{matrix} α C_{cer} (t) \cdot e^{- β t} f o r α e^{- β t} < 1 \\ C_{cer} (t) f o r α e^{- β t} > 1 \end{matrix}}

(2)

Coefficients α and β were estimated by fitting the upper part of this equation to the theoretical non-displaceable reference curve for each subject for t > 5 min, where it follows an exponential decay, and subsequently using their median values.

Convolution subtraction of an exponential function from the cerebellar TAC, as used by Johansson et al.¹:

B : C_{ref} (t) = C_{cer} (t) - 0.01 \cdot C_{cer} (t) \otimes e^{- 0.01 t}

(3)

Multiplication of the cerebellar TAC with a sigmoid function:

C : C_{ref} (t) = C_{cer} (t) \cdot \frac{(1 - γ \cdot t^{δ})}{t^{(δ + ε)}}

(4)

Here, $γ, δ and ε$ were estimated by fitting equation (4) to the non-displaceable reference curve for each subject. Again, the median value of each fitting parameter for the entire dataset was subsequently used.

To assess the general applicability of these median values, differences between parameter values across groups were analysed using 95% confidence intervals (CIs). A Kruskal–Wallis test (p < 0.05) was performed to determine whether significant differences existed between the groups, followed by post hoc Mann–Whitney U tests when a significant difference was found, with Bonferroni correction applied to account for multiple comparisons.

Parametric images

Parametric [¹¹C]L-deprenyl-D2 images, showing λk₃ at the voxel level, were generated using a basis function implementation⁴⁴ of the 2T3k plasma-input model and RRTM. For this analysis, 100 basis functions were predefined, with a discrete spectrum of parameter values for the exponential clearance rate constants k₂ + k₃ ranging between 0.015 and 1 min^–1. Furthermore, K_ND was estimated at the voxel level using the reference Patlak method with the various corrected cerebellar TACs as reference at the 10–60 min time interval. SBI images were generated as described above. Parametric methods were validated by comparison of regionally averaged λk₃ and K_ND values from the parametric images to VOI-based λk₃ and K_ND values using Spearman’s correlation and Deming regression.

A voxel-wise comparison between the MS groups (SPMS and AHSCT) and the healthy controls was performed using SPM12. First, all parametric images were smoothed with an 8 mm Gaussian filter and resampled into 2 mm isotropic voxels. Then, the co-registered 3D T1-weighted MR images were normalised to Montreal Neurological Institute (MNI) standard space, and the resulting transformation matrices were applied to respective parametric images. Mean parametric images of each group were compared using an independent two-sample t-test (t-maps were thresholded at clusters >1 cm³, and p < 0.01 (uncorrected) was considered significant).

Simulations

Simulations were performed to assess the identifiability of 2T3k K_ND and the implication of violation of the assumption of equal V_ND in target and reference tissues. Target and reference (without specific binding) TACs were simulated using 2T3k and 1T2k models, respectively, with a plasma input function based on the blood data of an arbitrarily chosen subject. A total of 100 TACs were generated, with rate constants randomly sampled from ranges derived from clinical data (K₁ = 0.4–0.7 ml.cm^–3.min^–1, k₂ = 0.1–0.2 min^–1, k₃ = 0.03–0.1 min^–1). The reference tissue K₁ was randomly distributed within ±0.1 ml.cm^–3.min^–1 of the target tissue K₁. Gaussian noise was added with a noise level mimicking typical levels in both reference and target tissues. Two scenarios were simulated: (1) k₂ was set to result in equal V_ND in reference and target tissues and (2) with differences between target and reference tissue V_ND by randomly varying reference tissue K₁ and k₂ within ±15% of the values of scenario 1.

Statistics

Statistical analysis was done as described in each section above. All kinetic modelling, parametric image generation and simulations were performed with in-house written software in MATLAB. Statistical analysis was performed in MATLAB and GraphPad Prism 10 (GraphPad Software, Boston, MA).

Results

Figure 1(a) shows representative TACs in striatum, thalamus, global cortical region and cerebellum. Figure 1(b) and (c) show the different modifications of the cerebellar TAC alongside the unmodified cerebellar curve and the theoretical non-displaceable reference curve based on K₁ and k₂ of the 2T3k plasma-input model. For modification A′, based on equation (2), the median values of the coefficients were: $α$ = 1.39 (95% CI 1.32–1.46) and $β$ = 0.059 (95% CI 0.058–0.063). For modification C, based on equation (4), the fitted parameters differed significantly between the AHSCT group and the groups with SPMS patients and healthy controls. Nevertheless, for consistency, median values across all subjects were used: $γ$ = 0.98 (95% CI 0.97–0.99), $δ$ = 2.72 (95% CI 2.64–2.82) and $ε$ = 2136 (95% CI 1482–3907).

Figure 1.

Time-activity curves for a representative subject; (a) striatum, thalamus, a global cortical region and CER; (b) uncorrected cerebellar TAC and modified cerebellar TACs using the previously suggested correction methods (A, B); (c) modified cerebellar TACs correction according to methods proposed in the current work (A′, C). The theoretical reference curve representing the CER-ND is shown in both (b) and (c).

Figure 2 shows the relationship between VOI-based λk₃ and K_ND values estimated with the 2T3k plasma-input model and reference Patlak using the uncorrected cerebellum TAC, as well as the previously proposed cerebellar TAC corrections A and B. For 2T3k K_ND, a moderate correlation with λk₃ was obtained in the striatum and thalamus, while a stronger correlation was observed in the global cortical region. Reference Patlak K_ND values, based on the corrected cerebellum TACs according to method A, showed a similar relation to λk₃ as 2T3k K_ND, but with higher correlations in subcortical regions than in the global cortical region. In contrast, reference Patlak K_ND values estimated with method B exhibited very poor correlation with λk₃, as seen with the uncorrected cerebellum TAC. Spearman correlation coefficients of reference Patlak K_ND values versus λk₃ are shown in Table 1. V_ND values in striatum and cerebellum, calculated using the 2T3k plasma-input model, showed a moderate correlation (ρ = 0.86; 95% CI 0.71–0.93) and a mean relative difference of 1% ± 11%.

Figure 2.

Plasma-input 2T3k λk₃ values in striatum, thalamus and a global cortical region versus K_ND values based on plasma-input 2T3k (a), Patlak analysis with the uncorrected CER TAC as reference (b), and Patlak analysis with modified CER reference TACs according to previously published methods A (c) and B (d).

Table 1.

Spearman correlation (ρ) between λk₃ estimates from plasma-input 2T3k and K_ND estimates from non-invasive methods with different cerebellar corrections, as well as SBI for the different time intervals. Data in parentheses represent 95% confidence intervals. The significance level was <0.0001 for overall correlations, except for the correlation between 2T3k λk₃ and reference Patlak K_ND without any correction of the cerebellar TAC (p = 0.0066).

Model	Striatum	Thalamus	Global	Overall
2T3k K_ND	0.49 (0.14–0.74)	0.35 (−0.04–0.65)	0.74 (0.49–0.87)	0.69 (0.56–0.79)
Patlak K_ND-CER	−0.47 (−0.72 to 0.10)	−0.50 (−0.74 to 0.15)	−0.26 (−0.58 to 0.14)	0.29 (0.08–0.48)
Patlak K_ND-A	0.68 (0.40–0.84)	0.73 (0.48–0.87)	0.53 (0.19–0.76)	0.82 (0.74–0.88)
Patlak K_ND-B	−0.04 (−0.42 to 0.35)	−0.14 (−0.50 to 0.26)	−0.13 (−0.49 to 0.26)	0.47 (0.28–0.62)
Patlak K_ND-CER-ND	0.65 (0.36–0.83)	0.63 (0.33–0.82)	0.51 (0.16–0.75)	0.83 (0.75–0.89)
Patlak K_ND-A′	0.72 (0.46–0.86)	0.79 (0.58–0.90)	0.57 (0.24–0.78)	0.84 (0.77–0.90)
Patlak K_ND-C	0.72 (0.46–0.86)	0.78 (0.57–0.90)	0.58 (0.25–0.79)	0.85 (0.77–0.90)
SBI_{30–40 min}	0.69 (0.42–0.85)	0.75 (0.51–0.88)	0.59 (0.26–0.79)	0.82 (0.74–0.88)
SBI_{50–60 min}	0.69 (0.42–0.85)	0.74 (0.50–0.88)	0.59 (0.26–0.79)	0.82 (0.74–0.88)

Figure 3 shows the relationship between plasma-input 2T3k λk₃ values and reference Patlak K_ND values based on the subject-specific non-displaceable reference TAC, the modified cerebellar corrections for [¹¹C]L-deprenyl-D2 binding A′ and C, as well as SBI_{50–60 min}. The corresponding Spearman correlation coefficients are provided in Table 1. As for 2T3k K_ND and method A, moderate correlations were generally found between plasma-input 2T3k λk₃ and reference Patlak K_ND across all VOIs and correction methods. Correlations were higher in the subcortical regions than in the global cortical region. The highest correlation between K_ND obtained with reference tissue models and plasma-input 2T3k λk₃ was observed with correction method C, though only slightly higher than for methods A and A′. As mentioned above, the fit parameters for method C differed significantly between groups, so group-specific reference TAC correction parameters were also tested for this method. This did not improve the overall correlation, which remained identical to that obtained using the median parameter values derived from all subjects (data not shown). Use of the non-displaceable cerebellum TAC as reference curve provided comparable results as methods A, A′ and C. Results obtained using RRTM with various correction methods were nearly identical to those found for the reference Patlak methods (data not shown). As seen in Figure 3(d) and Table 1, SBI shows a similar correlation to 2T3k λk₃ as reference Patlak K_ND using the best-performing cerebellar correction methods (A′ and C). The correlation between SBI and plasma-input 2T3k λk₃ was identical for both time intervals: 30–40 and 50–60 min.

Figure 3.

Plasma-input 2T3k λk₃ values in striatum, thalamus and a global cortical region versus K_ND values based on reference Patlak analysis with a subject-specific theoretical CER-ND TAC based on K₁ and k₂ of the 2T3k model (a), using the modified CER reference TACs according to methods A′ (b) and C (c), and versus a SBI defined as the ratio of target tissue uptake at 50–60 min p.i. to peak CER uptake (d).

Supplementary Figure S1 shows that the relations between the plasma-input 2T3k K_ND values and non-invasive Patlak K_ND for the non-displaceable cerebellum reference curve and the different correction methods, as well as between SBI_{50–60 min} and 2T3k K_ND, are similar with the exception of method B. Reference tissue corrections A′ and C show moderate correlations with slopes close to unity, and A shows this to a lesser extent. Reference Patlak K_ND based on method B correlates poorly with 2T3k K_ND. Spearman correlations and Deming regression slopes are given in Supplementary Table S1.

Figure 4(a) shows parametric [¹¹C]L-deprenyl-D2 λk₃, reference Patlak K_ND with different cerebellar correction methods, and SBI_{50–60 min} images for a representative subject with SPMS. Visually, the SBI and K_ND images show a similar pattern to the λk₃ image, albeit with cerebellar correction method B resulting in an underestimation of K_ND compared to the other methods, as expected based on Figure 2(d). As shown in Figure 4(b), VOI-based and voxel-based 2T3k λk₃ values demonstrated consistently high correlation and agreement. Correlation was particularly strong in the striatum (ρ = 0.96; 95% CI 0.90–0.98, slope = 0.90; 95% CI 0.78–1.02), and remained high in the thalamus (ρ = 0.92; 95% CI 0.83–0.96, slope = 0.86; 95% CI 0.71–1.01) and global cortical region (ρ = 0.92; 95% CI 0.83–0.96, slope = 0.96; 95% CI 0.61–1.30). These findings reflect an overall strong correlation across all regions (ρ = 0.97; 95% CI 0.95–0.98, slope = 0.86; 95% CI 0.80–0.93). Very high correlation and agreement were also observed between VOI-based and voxel-based reference Patlak K_ND values, with Spearman correlation coefficients and slopes close to unity.

Figure 4.

(a) Parametric [¹¹C]L-deprenyl-D2 images for a representative subject with SPMS (45 years, female). The top left shows the plasma-input 2T3k λk₃ image and the bottom left the SBI_{50–60 min} image. On the right, the reference Patlak K_ND images generated using the four different cerebellar correction methods are shown. (b) Relation between voxel-based and VOI-based 2T3k λk₃. The dashed line represents the line of identity.

Voxel-wise comparisons between SPMS patients and healthy controls based on the different models are presented in Figure 5. Both invasive and non-invasive methods showed a significant increase in MAO-B activity in white matter in SPMS patients compared to healthy controls. However, the reference Patlak methods and SBI images overestimated the extent of the difference between SPMS patients and healthy controls. No significant differences were found between the AHSCT group and healthy controls for plasma-input 2T3k λk₃ values. However, significant differences were observed in K_ND values with the modified reference Patlak methods (data not shown).

Figure 5.

Voxel-wise comparison between patients with SPMS and healthy controls. T-score maps display clusters with significantly higher 2T3k λk₃ and K_ND in (a) and (b), and SBI_{50–60 min} values in (c), indicating increased MAO-B activity in SPMS patients. Reference Patlak K_ND values with cerebellar correction A, B and C are shown in (d)–(f).

Table 2 shows the total volumes of the areas with increased MAO-B activity seen in Figure 5, along with mean and maximum significance levels, quantitatively illustrating the overestimation observed with the reference methods compared to plasma-input 2T3k λk₃. The reference Patlak methods consistently yielded volume estimates more than 10 times larger than those found with plasma-input 2T3k λk₃ for regions exceeding the significance threshold (t = 2.55, p < 0.01), along with elevated t-values within these clusters.

Table 2.

Volume of t-map clusters presented for plasma-input 2T3k λk₃ and K_ND estimates, as well as for SBI_{50–60 min} and the different reference Patlak methods, along with their corresponding mean and maximum t-values within these clusters.

Model	Volume (cm³)	Mean t-value	Max t-value
2T3k λk₃	32	3.12	5.23
2T3k K_ND	155	3.47	7.06
SBI_{50–60 min}	381	3.46	6.48
Patlak K_ND-A	527	4.31	9.94
Patlak K_ND-B	454	3.20	5.66
Patlak K_ND-C	483	4.15	9.45

The results of the simulations are shown in Supplementary Figure S2. Fitted 2T3k K_ND values showed a high correlation and agreement with the true values (ρ = 0.99, slope 1.01), indicating high accuracy and precision of 2T3k K_ND for noise levels corresponding to those at the VOI level. When assuming identical V_ND in target and reference tissue, reference Patlak K_ND also correlated and agreed well with 2T3k K_ND (ρ = 1.00, slope 1.02). However, when inducing a variation between target and reference tissue V_ND comparable to that found in the clinical data, the correlation between reference Patlak K_ND and 2T3k K_ND decreased to comparable levels as those seen in Supplementary Figure S1 (ρ = 0.87, slope 1.14), which explains the poor to moderate correlations between reference tissue and plasma-input models found in this work.

Discussion

In this study, we demonstrate that the implementation of parametric 2T3k λk₃ images with arterial input function using the basis function method agreed well with VOI-based weighted nonlinear regression analysis. Non-invasive methods showed moderate overall correlation with plasma-input λk₃ for the best-performing modified reference and semi-quantitative methods, with lower correlations for individual VOIs, such as the striatum, thalamus and a global cortical region, indicating that they are not a robust alternative to quantifying MAO-B activity using [¹¹C]L-deprenyl-D2. To the best of our knowledge, neither the parametric nor non-invasive methods have previously been thoroughly validated for [¹¹C]L-deprenyl-D2.

The net uptake rate relative to the non-displaceable volume of interest K_ND, even when based on arterial input data, did not show a high correlation with λk₃ (Figure 2(a)), in the global cortical but especially in the subcortical regions with high levels of MAO-B. This in itself is not surprising, as λk₃ and K_ND are different measures of tracer binding with different relationships to the irreversible binding rate constant k₃. λk₃ is proportional to k₃ provided V_ND (= λ) is constant, whereas K_ND is proportional to k₃ if k₃ ≪ k₂, which is somewhat fulfilled for [¹¹C]L-deprenyl-D2, with a decreasing slope in the relationship between K_ND and k₃ for increasing k₃ values relative to k₂. However, as λk₃ was previously suggested as the preferred measure of [¹¹C]L-deprenyl-D2 binding, we chose to use this as the gold standard value in the present work.

The two previously suggested non-invasive quantification methods, accounting for MAO-B binding in the cerebellar TAC either by multiplication with an exponential function (A)¹⁶ or by convolution subtraction of an exponential function (B),¹ did not provide K_ND values that correlated well with λk₃ either, but also not with 2T3k K_ND (Supplementary Figure S1). Efforts to optimise the previously proposed method A (equation (1)) by basing its parameters on a fit to the theoretical non-displaceable reference TAC (A′), or use of a sigmoid function to improve the reference TAC’s similarity to the theoretical non-displaceable reference TAC (C) (Figure 1), did not result in much improved agreement with the plasma-input model. For methods A′ and C, we used the median values of the fitted parameters from all subjects. Although we found a significant difference in parameters for method C between subject groups, the use of group-specific parameters did not improve overall correlation compared to using the median parameters derived from all subjects. However, our aim was to establish correction methods with parameters applicable to any subject, regardless of disease status, and implementing corrections tailored to specific patient groups would defeat this purpose.

The convolution subtraction correction method B (equation (3)) exhibited a very weak and negative correlation between K_ND and λk₃. Considering the shape of the reference curve estimated by this method, as seen in Figure 1(b), which does not correspond to the expected TAC for a region with only non-displaceable binding, this is not surprising. As mentioned in the introduction, this method has been previously used in a number of studies of neurodegenerative diseases.^{11–14,31–33} However, based on the present results, K_ND based on method B does not reflect MAO-B binding. Hence, it remains to be investigated how the results of those studies should be interpreted if they do not reflect MAO-B activity.

Even the use of the subject-specific theoretical non-displaceable cerebellum curve itself as reference TAC did not result in a strong correlation with 2T3k K_ND. One explanation for this could be that the non-displaceable cerebellum curve is not well defined, but the simulations showed high accuracy and precision of not only 2T3k K_ND but also of k₂ and k₃ themselves (data not shown). This is further supported by a test-retest study, which demonstrated that k₃ has similar reproducibility to λk₃ in healthy subjects.⁴⁵ However, k₃ is not a valid option for parametric images, as the noise level of individual voxel TACs prohibits robust estimation of k₃. Nevertheless, the simulations show that the moderate correlation of Patlak K_ND, using the non-displaceable cerebellum TAC as reference, with 2T3k K_ND is likely due to differences in V_ND between target and reference tissues rather than the robustness of 2T3k K_ND. Hence, even the existence of a reference region completely devoid of specific binding would not have resulted in high correlations between non-invasive K_ND and 2T3k K_ND in the presence of differences in V_ND between target and reference tissues as observed in the clinical data. This underscores the difficulty of achieving a fully reliable non-invasive alternative, as irreversible reference tissue models seem particularly sensitive to differences in V_ND between target and reference tissues.

Although correlations between K_ND and λk₃ are moderate at best using the different approaches for correction of the cerebellar TAC, the non-invasive K_ND images as well as SBI do visually resemble the λk₃ image as shown in Figure 4(a). However, as presented in Table 2, K_ND parametric images and SBI overestimate both the extent and magnitude of the significant difference between SPMS patients and healthy controls at the voxel level (Figure 5). Still, the visual resemblance between non-invasive parametric images and λk₃ images suggests that, although absolute quantification is challenging, they may be useful for assessment of intra-individual qualitative changes in MAO-B. If non-invasive methods are to be used, SBI provides a simple, easily implemented quantitatively comparable alternative to the best-performing modified reference tissue methods, while also enabling shorter scan duration if the 30–40 min interval is used.

In conclusion, parametric images of λk₃ obtained using a basis function implementation of the plasma-input irreversible two-tissue compartment model showed good agreement with the results from VOI-based nonlinear regression analysis. Full kinetic analysis with arterial input function is required for accurate assessment of MAO-B activity with [¹¹C]L-deprenyl-D2. Finally, this study highlights the importance of thorough validation of simplified methods prior to their use in clinical studies.

Supplemental Material

sj-docx-1-jcb-10.1177_0271678X251384264 – Supplemental material for Non-invasive modelling and parametric methods for quantification of MAO-B activity using [11C]L-deprenyl-D2 PET

Supplemental material, sj-docx-1-jcb-10.1177_0271678X251384264 for Non-invasive modelling and parametric methods for quantification of MAO-B activity using [11C]L-deprenyl-D2 PET by Karolina Hedman, My Jonasson, Lieuwe Appel, Andreas Tolf, Joachim Burman and Mark Lubberink in Journal of Cerebral Blood Flow & Metabolism

Supplemental Material

sj-docx-2-jcb-10.1177_0271678X251384264 – Supplemental material for Non-invasive modelling and parametric methods for quantification of MAO-B activity using [11C]L-deprenyl-D2 PET

Supplemental material, sj-docx-2-jcb-10.1177_0271678X251384264 for Non-invasive modelling and parametric methods for quantification of MAO-B activity using [11C]L-deprenyl-D2 PET by Karolina Hedman, My Jonasson, Lieuwe Appel, Andreas Tolf, Joachim Burman and Mark Lubberink in Journal of Cerebral Blood Flow & Metabolism

Footnotes

Acknowledgements

The authors would like to acknowledge the staff of the PET Centre at Uppsala University Hospital for their assistance with PET scans and tracer production, and the subjects for participating in the study.

Author contributions

Study design and ethical application: ML, JB and LA. Funding acquisition: JB. Recruitment of subjects and data acquisition: AT and LA. Data analysis and interpretation: KH, ML and MJ. First draft of the manuscript: KH. Continuous feedback during manuscript development: ML and MJ. Manuscript review: all authors. Approval of the final manuscript: all authors.

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: The study was supported by the Agreement on Medical Training and Clinical Research (ALF) for Uppsala County and grants from the Marianne and Marcus Wallenberg Foundation, the Swedish Foundation for MS Research, the Swedish Society for Medical Research and an unrestricted research grant from BiogenIdec Sweden AB.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

ORCID iDs

Karolina Hedman

My Jonasson

Andreas Tolf

Joachim Burman

Mark Lubberink

Data availability statement

Data is available upon reasonable request by contacting the authors. Raw data (PET images, etc.) can only be made available under the condition of appropriate ethics approval.

Supplemental material

Supplemental material for this article is available online.

References

Johansson

Engler

Blomquist

, et al. Evidence for astrocytosis in ALS demonstrated by [¹¹C](L)-deprenyl-D₂ PET. J Neurol Sci 2007; 255: 17–22.

Jacobs

Tavitian

Noninvasive molecular imaging of neuroinflammation. J Cereb Blood Flow Metab 2012; 32: 1393–1415.

Gulyás

Pavlova

Kása

, et al. Activated MAO-B in the brain of Alzheimer patients, demonstrated by [¹¹C]-L-deprenyl using whole hemisphere autoradiography. Neurochem Int 2011; 58: 60–68.

Ekblom

Jossan

Bergstrüm

, et al. Monoamine oxidase-B in astrocytes. Glia 1993; 8: 122–132.

Magyar

(−)-Deprenyl, a selective MAO-B inhibitor, with apoptotic and anti-apoptotic properties. Neurotoxicology 2004; 25: 233–242.

Knoll

Magyar

Some puzzling pharmacological effects of monoamine oxidase inhibitors. Adv Biochem Psychopharmacol 1972; 5: 393–408.

Fowler

MacGregor

Wolf

, et al. Mapping human brain monoamine oxidase A and B with ¹¹C-labeled suicide inactivators and PET. Science 1987; 235: 481–485.

Fowler

Wang

Logan

, et al. Selective reduction of radiotracer trapping by deuterium substitution: comparison of carbon-11-L-deprenyl and carbon-11-deprenyl-D₂ for MAO B mapping. J Nucl Med 1995; 36: 1255–1262.

Arnett

Fowler

MacGregor

, et al. Turnover of brain monoamine oxidase measured in vivo by positron emission tomography using L-[¹¹C]deprenyl. J Neurochem 1987; 49: 522–527.

10.

Santillo

Gambini

Lannfelt

, et al. In vivo imaging of astrocytosis in Alzheimer’s disease: an ¹¹C-L-deuteriodeprenyl and PIB PET study. Eur J Nucl Med Mol Imaging 2011; 38: 2202–2208.

11.

Carter

Schöll

Almkvist

, et al. Evidence for astrocytosis in prodromal Alzheimer disease provided by ¹¹C-deuterium-L-deprenyl: a multitracer PET paradigm combining ¹¹C-Pittsburgh compound B and ¹⁸F-FDG. J Nucl Med 2012; 53: 37–46.

12.

Schöll

Carter

Westman

, et al. Early astrocytosis in autosomal dominant Alzheimer’s disease measured in vivo by multi-tracer positron emission tomography. Sci Rep 2015; 5: 16404.

13.

Rodriguez-Vieitez

Carter

Chiotis

, et al. Comparison of early-phase ¹¹C-deuterium-l-deprenyl and ¹¹C-Pittsburgh compound B PET for assessing brain perfusion in Alzheimer disease. J Nucl Med 2016; 57: 1071–1077.

14.

Chiotis

Johansson

Rodriguez-Vieitez

, et al. Tracking reactive astrogliosis in autosomal dominant and sporadic Alzheimer’s disease with multi-modal PET and plasma GFAP. Mol Neurodegener 2023; 18: 60.

15.

Kumlien

Nilsson

Hagberg

, et al. PET with ¹¹C-deuterium-deprenyl and ¹⁸F-FDG in focal epilepsy. Acta Neurol Scand 2001; 103: 360–366.

16.

Bergström

Kumlien

Lilja

, et al. Temporal lobe epilepsy visualized with PET with ¹¹C-L-deuterium-deprenyl – analysis of kinetic data. Acta Neurol Scand 1998; 98: 224–231.

17.

Kumlien

Bergström

Lilja

, et al. Positron emission tomography with [¹¹C]deuterium-deprenyl in temporal lobe epilepsy. Epilepsia 1995; 36: 712–721.

18.

Engler

Lundberg

Ekbom

, et al. Multitracer study with positron emission tomography in Creutzfeldt–Jakob disease. Eur J Nucl Med Mol Imaging 2003; 30: 85–95.

19.

Fowler

Comparison of brain glucose metabolism and monoamine oxidase B (MAO B) in traumatic brain injury. Clin Positron Imaging 1999; 2: 71–79.

20.

Bench

Price

Lammertsma

, et al. Measurement of human cerebral monoamine oxidase type B (MAO-B) activity with positron emission tomography (PET): a dose ranging study with the reversible inhibitor Ro 19-6327. Eur J Clin Pharmacol 1991; 40: 169–173.

21.

Hirvonen

Kailajärvi

Haltia

, et al. Assessment of MAO-B occupancy in the brain with PET and [¹¹C]-L-deprenyl-D₂: a dose-finding study with a novel MAO-B inhibitor, EVT 301. Clin Pharmacol Ther 2009; 85: 506–512.

22.

Fowler

Volkow

Logan

, et al. Monoamine oxidase B (MAO B) inhibitor therapy in Parkinson’s disease. Neurology 1993; 43: 1984–1984.

23.

Sturm

Forsberg

Nave

, et al. Positron emission tomography measurement of brain MAO-B inhibition in patients with Alzheimer’s disease and elderly controls after oral administration of sembragiline. Eur J Nucl Med Mol Imaging 2016; 44: 382.

24.

Reich

Lucchinetti

Calabresi

PA.

Multiple sclerosis. N Engl J Med 2018; 378: 169–180.

25.

Janssen

Vugts

Funke

, et al. Imaging of neuroinflammation in Alzheimer’s disease, multiple sclerosis and stroke: recent developments in positron emission tomography. Biochim Biophys Acta 2016; 1862: 425–441.

26.

Brownlee

Hardy

Fazekas

, et al. Diagnosis of multiple sclerosis: progress and challenges. Lancet 2017; 389: 1336–1346.

27.

Lammertsma

Bench

Price

, et al. Measurement of cerebral monoamine oxidase B activity using L-[¹¹C]deprenyl and dynamic positron emission tomography. J Cereb Blood Flow Metab 1991; 11: 545–556.

28.

Logan

Fowler

Volkow

, et al. Reproducibility of repeated measures of deuterium substituted [¹¹C]L-deprenyl ([¹¹C]L-deprenyl-D₂) binding in the human brain. Nucl Med Biol 2000; 27: 43–49.

29.

Fowler

Logan

Volkow

, et al. Translational neuroimaging: positron emission tomography studies of monoamine oxidase. Mol Imaging Biol 2005; 7: 377–387.

30.

Fowler

Logan

Volkow

, et al. Monoamine oxidase: radiotracer development and human studies. Methods 2002; 27: 263–277.

31.

Carter

Chiotis

Nordberg

, et al. Longitudinal association between astrocyte function and glucose metabolism in autosomal dominant Alzheimer’s disease. Eur J Nucl Med Mol Imaging 2019; 46: 348–356.

32.

Rodriguez-Vieitez

Saint-Aubert

Carter

, et al. Diverging longitudinal changes in astrocytosis and amyloid PET in autosomal dominant Alzheimer’s disease. Brain 2016; 139: 922–936.

33.

Murugan

Chiotis

Rodriguez-Vieitez

, et al. Cross-interaction of tau PET tracers with monoamine oxidase B: evidence from in silico modelling and in vivo imaging. Eur J Nucl Med Mol Imaging 2019; 46: 1369–1382.

34.

Tolf

Fagius

Carlson

, et al. Sustained remission in multiple sclerosis after hematopoietic stem cell transplantation. Acta Neurol Scand 2019; 140: 320–327.

35.

Fowler

Wolf

MacGregor

, et al. Mechanistic positron emission tomography studies: demonstration of a deuterium isotope effect in the monoamine oxidase-catalyzed binding of [¹¹C]-L-deprenyl in living baboon brain. J Neurochem 1988; 51: 1524–1534.

36.

Macgregor

Fowler

Wolf

, et al. Synthesis of suicide inhibitors of monoamine oxidase: carbon-11 labeled clorgyline, L-deprenyl and D-deprenyl. J Label Compd Radiopharm 1988; 25: 1–9.

37.

Brix

Zaers

Adam

, et al. Performance evaluation of a whole-body PET scanner using the NEMA protocol. National Electrical Manufacturers Association. J Nucl Med 1997; 38(10): 1614–1623.

38.

Hsu

DFC

Ilan

Peterson

, et al. Studies of a next-generation silicon-photomultiplier-based time-of-flight PET/CT system. J Nucl Med 2017; 58: 1511–1518.

39.

Svarer

Madsen

Hasselbalch

, et al. MR-based automatic delineation of volumes of interest in human brain PET images using probability maps. Neuroimage 2005; 24: 969–979.

40.

Thiele

Buchert

Evaluation of non-uniform weighting in non-linear regression for pharmacokinetic neuroreceptor modelling. Nucl Med Commun 2008; 29: 179–188.

41.

Patlak

Blasberg

Fenstermacher

JD.

Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. J Cereb Blood Flow Metab 1983; 3: 1–7.

42.

Patlak

Blasberg

RG.

Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. Generalizations. J Cereb Blood Flow Metab 1985; 5(4): 584–590.

43.

Lammertsma

Bench

Hume

, et al. Comparison of methods for analysis of clinical [¹¹C]raclopride studies. J Cereb Blood Flow Metab 1996; 16: 42–52.

44.

Gunn

Lammertsma

Hume

, et al. Parametric imaging of ligand-receptor binding in PET using a simplified reference region model. Neuroimage 1997; 6: 279–287.

45.

Arakawa

Stenkrona

Takano

, et al. Test–retest reproducibility of [¹¹C]-l-deprenyl-D₂ binding to MAO-B in the human brain. EJNMMI Res 2017; 7: 54.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.03 MB