First-in-human neuroimaging of soluble epoxide hydrolase using [¹⁸F]FNDP PET

Abstract

Soluble epoxide hydrolase (sEH) is an enzyme with putative effect on neuroinflammation through its influence on the homeostasis of polyunsaturated fatty acids and related biproducts. Since sEH is an enzyme that metabolizes the anti-inflammatory epoxy fatty acids to relatively inert 1,2-diol forms, a high availability or activity of sEH may promote vasoconstriction and inflammation in local tissues that may be linked to neuropsychiatric disease such as depressive disorders. [¹⁸F]FNDP is a radiotracer developed by our group for studying sEH in the living human brain with positron emission tomography (PET).

Methods:

Brain PET using bolus injection of [¹⁸F]FNDP followed by emission imaging lasting 90 or 180 min was completed in seven healthy adults (5 males, 2 females, ages 40–53 years). The kinetic behavior of [¹⁸F]FNDP was evaluated using a metabolite corrected arterial plasma input function with compartmental or graphical modeling approaches.

Results:

[¹⁸F]FNDP PET was well tolerated. Akaike information criterion favored the two-tissue compartment model (2TCM) in all ten regions of interest. Regional total distribution volume (V_T) values from each compartmental model and Logan analysis were generally well-identified except corpus callosum V_T using 2TCM. Logan analysis was assessed as the choice model due to stability of the regional V_T values from 90 min data and due to high correlation of Logan-derived regional V_T values with those from the 2TCM. The distribution of sEH differed from that of our prior study of primate brain, with higher binding in human cerebellar cortex and thalamus relative to other cortical regions, which aligns with reported consensus expression patterns of the epoxide hydrolase 2 gene in human brain.

Conclusion:

These data support further use of [¹⁸F]FNDP PET to study sEH in human brain.

Introduction

Soluble epoxide hydrolase (sEH) is an enzyme expressed throughout the body, and plays a role in maintaining homeostasis of polyunsaturated fatty acids and their biproducts. This homeostasis may have effect on immune signaling, as reviewed by Zarriello et al (1). For example, polyunsaturated fatty acids can be metabolized to epoxy-fatty acids (epoxyeicosatrienoic or epoxydocosapentaenoic acids) that promote vasodilatation and anti-inflammatory pathways, while the metabolism of polyunsaturated fatty acids through the cyclooxygenase pathway yields more inflammatory derivatives (prostaglandins). Since sEH is an enzyme that metabolizes anti-inflammatory epoxy fatty acids to relatively inert 1,2-diol forms, a high availability or activity of sEH may promote vasoconstriction and inflammation in local tissues (1). This inflammatory microenvironment may foster and/or exacerbate pathologic changes in several clinical conditions (2,3), and therapies to block sEH activity are therefore in various stages of development (1).

We developed N-(3,3-diphenylpropyl)-6-¹⁸F-fluronicotinamide ([¹⁸F]FNDP) for use with positron emission tomography (PET) to study further the proposed link between high availability of sEH in the human brain and neurologic [Lewy body dementia (3)] or psychiatric [affective and psychotic (2)] conditions. [¹⁸F]FNDP PET may also provide utility in assessing target engagement of drugs developed to block sEH activity in brain. [¹⁸F]FNDP showed high binding specificity in CD-1 and sEH knockout mice (4). In brains of Papio Anubis baboons, [¹⁸F]FNDP demonstrated reversible radiotracer pharmacokinetics and [¹⁸F]FNDP binding [total distribution volume (V_T)] was estimated well in 18 regions of interest (ROIs) using 90 min dynamic emission data with a metabolite corrected arterial input function (4). Radiometabolite analyses in the baboon studies demonstrated evidence of two relatively hydrophilic metabolites of [¹⁸F]FNDP. In baboon brains, [¹⁸F]FNDP V_T was highest in insula, putamen, caudate, and amygdala, and intermediate in frontal and temporal cortices, as well as hippocampus (4). Globus pallidus, white matter, hypothalamus, and thalamus had relatively lower [¹⁸F]FNDP V_T estimates, and lowest [¹⁸F]FNDP V_T occurred in cerebellum (4). Pre-PET pharmacological treatment with a sEH inhibitor (nor-fluoro-FNDP) resulted in 95% blockade of the [¹⁸F]FNDP binding in baboon brain (4), supporting binding specificity.

Here we report the first use of [¹⁸F]FNDP PET in the healthy human brain. Since there are no known brain regions without soluble epoxide hydrolase in humans, quantitative binding was estimated using [¹⁸F]FNDP imaging data with arterial input function.

Materials and Methods

Human Subject participation

This prospective study was approved by a Johns Hopkins Institutional Review Board. Healthy individuals between ages of 35–55 years were recruited through local advertising and provided written, informed consent. Each subject completed a screening interview and blood work, electrocardiogram, and urine toxicology. Eligible participants had stable health with no clinical abnormality on the screening assessment or structural magnetic resonance imaging (MRI). Exclusion criteria included past neurological or psychiatric illness, history of substance abuse including marijuana (assessed by self-report and urine toxicology), history of recent infection, history of head trauma with loss of consciousness, contraindication to MRI, or contraindication to PET with arterial line.

[¹⁸F]FNDP PET in humans

Synthesis and injection of [¹⁸F]FNDP

[¹⁸F]FNDP was synthesized as previously described (5) and radiochemical purity was 100% at the end of each synthesis.

Brain PET image acquisition

PET data were acquired on a CPS/CTI High Resolution Research Tomograph (CPS Innovations, Inc., Knoxville, TN). Prior to scanning, a thermoplastic mask was molded to the participant’s face for head fixation. A radial artery catheter was placed for arterial blood sampling, and an intravenous catheter was placed for radiotracer administration. A transmission scan was performed prior to the emission scan that began with bolus injection of [¹⁸F]FNDP. Continuous emission data were collected for 180 min (N = 3) or 90 min (N = 4) post injection (p.i.).

The PET data were binned into 30 frames (90 min scans) or 48 frames (180 min scans): 4 ×15 s, 4 × 30 s, 3 × 1 min, 2 × 2 min, 5 × 4 min, and 12 x or 30 × 5 min frames. These data were reconstructed using the iterative ordered subsets expectation maximization algorithm (6) (reconstructed image: 256 × 256 × 207 voxels, with voxel size: 1.22 × 1.22 × 1.22 mm³).

Plasma analysis

Approximately 30–35 arterial blood samples (each 1 mL) were collected p.i. at the following time points: as fast as possible over 1.5 min; every 30 s between 1.5–3 min; every 2 min between 3–5 min; and every 5 min between 5 min to the end of the emission scan. Plasma was isolated by centrifugation at 4 °C (5 min at 2200 g) and radioactivity was counted using a 2480 WIZARD² gamma counter (PerkinElmer, Waltham, MA).

Larger (~4–6 mL) blood samples for measurement of parent [¹⁸F]FNDP and metabolites occurred at the following time points: before [¹⁸F]FNDP injection for background measure, and p.i. at 5, 10, 20, 30, 45, 60, 75, and 90 min. Additional blood samples were obtained at 105, 120, 150, and 180 min p.i. for emission scans lasting 180 min (N=3). Parent [¹⁸F]FNDP and metabolites were analyzed using a high performance liquid chromatography (HPLC) system from Agilent Technologies consisting of a 1260 Infinity quaternary pump, a 1260 Infinity column compartment module, a 1260 Infinity UV detector, and Raytest GABI Star radiation detectors controlled by OpenLab CDS EZChrom (A.01.04) software. The HPLC system was first standardized using [¹⁸F]FNDP and a non-radioactive analog of the tracer prior to analysis. Plasma samples were loaded onto a 2 mL Rheodyne injector loop and directed to a capture column (packed with Phenomenex Strata-X 33μm polymeric reversed phase sorbent) with 1% acetonitrile and 99% water mobile phase at 2 mL/min. The effluent from the capture column contained polar metabolites measured by a detector. After 2 min of elution, an analytical mobile phase (65% acetonitrile, 35% aqueous 0.06 M ammonium formate) was applied to direct the trapped non-polar metabolites and [¹⁸F]FNDP to an analytical column (4.6 × 250 mm, XBridge Column, 5 μm) at 2 mL/min, which were then measured by a detector. HPLC chromatograms were integrated to provide percentage of parent [¹⁸F]FNDP relative to radiometabolite peaks at each time point. Metabolite-corrected plasma time-activity curves (TACs) were obtained by applying the percent parent [¹⁸F]FNDP time-profiles from HPLC to the total plasma TACs after linear interpolation using PMOD (v3.7, PMOD Technologies Ltd, Zurich, Switzerland).

Plasma free fraction (f_P) of [¹⁸F]FNDP was assessed using plasma isolated from blood sampling prior to radiotracer injection. Plasma (1 mL) was incubated with 0.37 MBq of [¹⁸F]FNDP and incubated (5 min) at room temperature. Then three separate 150 uL aliquots of the mixture were applied to Centrifree membrane filters (Millipore, Burlington, MA) and centrifuged (20 min, 26000 g, 25 °C). To calculate f_P, 50 uL samples of elute and three 50 uL samples of plasma incubated with [¹⁸F]FNDP were counted on an automated gamma counter, with f_P = counts in ultrafiltrate relative to plasma.

Brain MRI acquisition and regional segmentation

Prior to PET, each participant completed a sagittal T1-weighted brain MRI sequence acquired using a 3 Tesla Siemens MAGNETOM Prisma scanner (Malvern, PA, USA). Regions of interest (ROIs) were segmented from magnetization-prepared rapid gradient-echo data (spatial resolution 0.8 × 0.8 × 0.8 mm³) using the FreeSurfer image analysis suite (http://surfer.nmr.mgh.harvard.edu/). The ten ROIs included a white matter region (corpus callosum), cortical regions (cerebellar, parietal, frontal, cingulate, occipital, and temporal cortices), as well as subcortical regions (thalamus, striatum, and hippocampus).

Kinetic analysis

PMOD was used for PET data processing and kinetic analyses, as well as pre-processing steps of motion correction and co-registration of PET and MRI data. Regional TACs were generated after applying the ROI template to the dynamic PET data that were transformed into MR space. Regional V_T (7) values were derived using the metabolite-corrected arterial input function with the one- or two-tissue compartment model (1TCM, 2TCM), or Logan graphical analysis (8). Logan-derived V_T values were first evaluated using unfixed t* that guided the choice of fixed t* =30 min. Cerebral blood volume was set at 5% of brain volume when accounting for activity in the vasculature.

Model choice (1TCM v. 2TCM) was evaluated based on goodness of fit using the Akaike Information Criterion (AIC). The identifiability of V_T was assessed using the percent standard error (% SE <10%) from the theoretical parameter covariance matrix. For the three individuals with 180 min scans, regional V_T values were evaluated using varied, shortened scan durations down to 90 min data. In these analyses, V_T values derived from 180 min data were used as the standard for comparison. For each shortened scan duration (X), the relative bias values were expressed as |V_{T X min} – V_{T 180 min}|/ V_{T 180 min}.

Results

Radiotracer injection and Plasma analysis

Seven participants (5 males, 2 females, ages 40–53 years, median 52 years) underwent [¹⁸F]FNDP PET (Table 1). Molar radioactivity was 1,591.9 ± 1,289.0.0 GBq/μmol at the time of injection. The mean administered mass and radioactivity of [¹⁸F]FNDP were 0.13 ± 0.09 μg (range 0.03 – 0.26 μg) and 355.7 ± 25.6 MBq (range 309.0 – 378.1 MBq), respectively. Injection of [¹⁸F]FNDP did not yield adverse or clinical pharmacologic effects and there were no significant changes in pre- versus post-injection laboratory (blood, electrocardiogram) results.

TABLE 1.

Clinical characteristics of 7 healthy human participants.

Age (years)	48.7 ± 4.9*
Sex (Male/Female)	5/2
Race (Caucasian/African American)	3/4
Body mass index	28.0 ± 3.2*

^*Presented as mean ± standard deviation

Plasma activity peaked within 90 s p.i. and decreased to < 5% of the peak by 5 min (Figure 1A for representative curve). The mean parent fraction at 180 min p.i. (N=3) was 16.6 ± 6.5%, and at shorter duration of 90 min p.i. (N=7, Fig 1B) was 27.2 ± 6.9%. f_P was 1.5 ± 0.2% (N=7).

FIGURE 1.

Plasma profile of parent [¹⁸F]FNDP and radiometabolites in healthy individuals. A) Total activity concentration in plasma and parent [¹⁸F]FNDP activity in plasma in a subject who underwent a 180 min emission scan. SUV is radioactivity concentration as a fraction of the injected dose per mL plasma, normalized to body weight in g. B) Fraction of parent [¹⁸F]FNDP in plasma over a 90 min emission scan (N=7), shown as mean ± standard deviation. C) Radio-HPLC chromatogram at 180 min post injection of [¹⁸F]FNDP, showing elution of : 1. a polar radiometabolite species not retained on the capture column (26%), 2. a radiometabolite that was retained on the capture column (58 %) and 3. parent [¹⁸F]FNDP (16%) that is more lipophilic than species #2 as demonstrated by ~2 min longer retention time.

Modeling results

[¹⁸F]FNDP uptake across the brain ROIs is shown in Fig 2. TACs in corpus callosum peaked earlier (~2.5–3.5 min) compared to other regions that peaked ~ 7 min. Highest peak radioactivity concentration (SUV) occurred in cerebellar cortex, and lowest SUV peak was observed in corpus callosum.

FIGURE 2.

Regional radioactivity time-activity curves (TACs) in a representative healthy human participant. TACs are shown as percentages of the injected dose per cm³ tissue normalized to body weight in grams (SUV).

Visually, the 1TCM and 2TCM each fit the regional data. AIC favored the 2TCM in all ten ROIs (Supplemental Table 1). Well identified rate constants and V_T values from each compartmental model and Logan analysis are summarized in Table 2. The 2TCM identified K₁ and V_T well for all ROIs except for V_T in corpus callosum that had poor identifiability in four of seven participants, as well as V_T in hippocampus and striatum in two separate individuals. Rate constants from the 2TCM other than K₁ were moderately or poorly identified across all ten ROIs. The 1TCM identified well K₁, k_2, and V_T for all ROIs. Using 90 min data (N=7), V_T values derived from each model (1TCM, 2TCM, Logan) were highest in cerebellar cortex, thalamus, striatum and hippocampus, with intermediate V_T values in other cortical ROIs, and lowest V_T in corpus callosum. After excluding data from corpus callosum due to the relative, poor identifiability of V_T, as well as the one case of poorly identified V_T in each of hippocampus and striatum, regional V_T values were well-correlated between each model: V_{T 2TCM} = 1.07 V_{T 1TCM} - 0.08, R² =0.99; V_{T Logan, t* 30 min} = 0.97 V_{T 2TCM} + 0.02, R² =0.99; V_{T Logan, t* 30 min} = 1.03 V_{T 1TCM} – 0.07, R² =0.99. Parametric images of Logan-derived V_T from 90 min emission data are shown in Fig 3.

TABLE 2.

Modeling of [¹⁸F]FNDP PET data from 90 min emission scans (N=7) identified well the kinetic parameters and total distribution volume (V_T) values using the one-tissue compartment model (1TCM) in all regions. K₁ and V_T values were well identified using a two-tissue compartment model (2TCM) in most regions, and V_T values using Logan analysis were well identified in all regions.

	1TCM			2TCM		Logan
ROI	K 1	k 2	VT	K 1	VT	VT
	(mL cm−3 min−1)	(min−1)	(mL cm−3)	(mL cm−3 min−1)	(mL cm−3)	(mL cm−3)
Thalamus	0.17 ± 0.02	0.03 ± 0.01	6.15 ± 0.83	0.19 ± 0.03	6.48 ± 0.93	6.30 ± 0.91
Striatum	0.17 ± 0.02	0.03 ± 0.01	5.45 ± 0.68	0.18 ± 0.03	5.72 ± 0.78*	5.49 ± 0.70
Hippocampus	0.13 ± 0.02	0.02 ± 0.00	5.57 ± 0.81	0.15 ± 0.03	5.98 ± 0.99*	5.79 ± 0.87
Corpus Callosum	0.07 ± 0.01	0.02 ± 0.00	2.96 ± 0.47	0.08 ± 0.02	3.24 ± 0.36†	3.34 ± 0.62
Cerebellar CTX	0.17 ± 0.02	0.03 ± 0.00	6.59 ± 0.87	0.20 ± 0.03	6.91 ± 0.94	6.70 ± 0.92
Temporal CTX	0.14 ± 0.02	0.03 ± 0.01	4.32 ± 0.65	0.15 ± 0.03	4.59 ± 0.73	4.41 ± 0.68
Occipital CTX	0.18 ± 0.02	0.04 ± 0.01	4.60 ± 0.67	0.19 ± 0.03	4.84 ± 0.71	4.73 ± 0.70
Cingulate CTX	0.17 ± 0.02	0.04 ± 0.01	4.64 ± 0.66	0.18 ± 0.03	4.88 ± 0.71	4.72 ± 0.72
Frontal CTX	0.16 ± 0.02	0.04 ± 0.01	4.37 ± 0.57	0.18 ± 0.03	4.54 ± 0.60	4.43 ± 0.61
Parietal CTX	0.17 ± 0.02	0.04 ± 0.01	4.54 ± 0.62	0.18 ± 0.03	4.74 ± 0.65	4.63 ± 0.63

Presented as mean ± standard deviation for each estimated parameter. Identifiability of each parameter was assessed as % standard error <10%, and when not met the value was excluded:

^*N=6,

^†N=2. Region of Interest, ROI; Cortex, CTX.

FIGURE 3. Mean parametric image map of the total distribution volume (VT) of [18F]FNDP.

For each of the healthy participants (N=7), a V_T parametric map was computed from the 90 min [¹⁸F]FNDP dynamic PET data using Logan analysis (t* = 30 min) with the metabolite corrected plasma input function. The V_T maps were then normalized to a SPM8 brain template and averaged to generate a mean V_T parametric map. V_T is in units of mL cm⁻³.

Time stability of V_T

Using data from the three healthy individuals with 180 min data, Logan-derived V_T values from data shortened to 90 min were within 4% of the V_T values obtained using the full 180 min data. The minimum scan time needed to achieve stability in Logan-derived V_T value for each ROI is listed in Supplemental Table 2.

Discussion

We present human neuroimaging data that support further use of [¹⁸F]FNDP PET to estimate availability of sEH, which may be locally elevated in neuropsychiatric disorders like affective illness or schizophrenia (2). [¹⁸F]FNDP PET was well-tolerated and peak brain radioactivity concentrations (SUV) ranged from ~2.4–3.4 for all ROIs except corpus callosum that had lower peak (SUV ~1.5).

Regional V_T values were identified well using each model except when using the 2TCM to estimate V_T in corpus callosum. The 2TCM was favored over the 1TCM by goodness of fit. Using 1TCM or 2TCM, K₁ was identified well in all ROIs, and ranged between 0.14–0.19 mL cm⁻³ min⁻¹ in all regions except corpus callosum that had lower K₁ (0.07 or 0.08 mL cm⁻³ min⁻¹ using 1TCM or 2TCM, respectively). Logan graphical analysis was selected as the choice model due to the relative stability of regional V_Ts derived from this graphical modeling approach, and due to the high correlation between Logan-derived regional V_Ts and those from the 2TCM. Logan-derived V_T values across all the ROIs achieved relative stability by 90 min p.i.

V_T values derived using each model were highest in cerebellar cortex, thalamus, and hippocampus (listed in sequentially decreasing order), across the ten ROIs that included five cortical brain regions. These results are generally consistent with reported, higher mRNA expression of the epoxide hydrolase 2 gene (EPHX2) in cerebellum, thalamus, and hippocampal formation compared to cerebral cortex (9), although the ROI definitions used in this study deviate somewhat from those used in the two transcriptomics datasets from which consensus normalized EPHX2 expression levels were derived (9) (for example, cerebellar cortex versus cerebellum). These human imaging data suggest a different pattern of sEH distribution in human brain compared to that of baboon, since we found evidence of low sEH in baboon cerebellum using [¹⁸F]FNDP PET (4).

Conclusion

[¹⁸F]FNDP PET is a promising technique for studying soluble epoxide hydrolase in human brain in vivo. These data support the use of Logan analysis to derive regional [¹⁸F]FNDP V_T values using 90 min emission data in healthy human brain.

KEY POINTS

QUESTION:

Does [¹⁸F]FNDP have suitable kinetic behavior in human brain to support its use to study soluble epoxide hydrolase in vivo with PET?

PERTINENT FINDINGS:

[¹⁸F]FNDP PET was well tolerated and showed a pattern of uptake consistent with fast, reversible kinetics that facilitated generation of stable binding outcomes in each brain region using Logan graphical analysis, 90 min emission data, and the metabolite-corrected plasma input function.

IMPLICATIONS FOR PATIENT CARE:

[¹⁸F]FNDP PET is a promising technique for studying the role of soluble epoxide hydrolase in human health and disease in vivo .