Quantification of Cardiac Sympathetic Nerve Density with N-¹¹C-guanyl-meta-Octopamine and Tracer Kinetic Analysis

Abstract

Most cardiac sympathetic nerve radiotracers are substrates of the norepinephrine transporter (NET). Existing tracers like ¹²³I-metaiodobenzylguanidine (¹²³I-MIBG) and ¹¹C-(−)-meta-hydroxyephedrine (¹¹C-HED) are ‘flow-limited’ tracers due to their rapid NET transport rates. This prevents successful application of kinetic analysis techniques and causes semi-quantitative measures of tracer retention to be insensitive to mild-to-moderate nerve losses. N-¹¹C-guanyl-(−)-meta-octopamine (¹¹C-GMO) has a much slower NET transport rate and is trapped in storage vesicles. The goal of this study was to determine if analyses of ¹¹C-GMO kinetics could provide robust and sensitive measures of regional cardiac sympathetic nerve densities.

Methods

PET studies were performed in a rhesus macaque monkey under control conditions or following intravenous infusion of the NET inhibitor desipramine (DMI). Five DMI dose levels were used to establish a range of available cardiac NET levels. Compartmental modeling of ¹¹C-GMO kinetics yielded estimates of the rate constants K₁ (mL/min/g), k₂ (min⁻¹), and k₃ (min⁻¹). These values were used to calculate a ‘net uptake rate’ constant K_i (mL/min/g) = (K₁k₃)/(k₂ + k₃). In addition, Patlak graphical analyses of ¹¹C-GMO kinetics yielded Patlak slopes K_p (mL/min/g), which represent alternative measurements of the net uptake rate constant K_i. ¹¹C-GMO kinetics in isolated rat hearts were also measured for comparison with other tracers.

Results

In isolated rat hearts, the neuronal uptake rate of ¹¹C-GMO was 8 times slower than ¹¹C-HED and 12 times slower than ¹¹C-MIBG. ¹¹C-GMO also had a very long neuronal retention time (>200 h). Compartmental modeling of ¹¹C-GMO kinetics in monkey heart proved stable under all conditions. Calculated net uptake rate constants K_i tracked DMI-induced reductions of available NET in a dose-dependent manner (IC₅₀ = 0.087 ± 0.012 mg/kg DMI). Patlak analysis provided highly linear Patlak plots and the Patlak slopes K_p also declined in a dose-dependent manner (IC₅₀ = 0.068 ± 0.010 mg/kg DMI).

Conclusion

Compartmental modeling and Patlak analysis of ¹¹C-GMO kinetics each provided quantitative parameters that accurately tracked changes in cardiac NET levels. These results strongly suggest that PET studies with ¹¹C-GMO can provide robust and sensitive quantitative measures of regional cardiac sympathetic nerve densities in human hearts.

INTRODUCTION

It has been more than thirty years since radioiodinated metaiodobenzylguanidine (MIBG) was first introduced for scintigraphic imaging of cardiac sympathetic innervation (1). Since then, several other radiotracers, including ¹¹C-(−)-meta-hydroxyephedrine (¹¹C-HED) and ¹¹C-(−)-epinephrine (¹¹C-EPI), have been developed to assess cardiac sympathetic nerve integrity with positron emission tomography (PET) (2). Clinical studies with these tracers have made significant contributions to our understanding of cardiac sympathetic dysfunction in many diseases (3). In addition, recent studies point to an emerging role for cardiac neuronal imaging in the identification of heart failure patients at high risk for sudden cardiac death (4).

As structural analogs of the neurotransmitter norepinephrine, these tracers are actively transported into neurons by the norepinephrine transporter (NET) and stored in norepinephrine storage vesicles by the vesicular monoamine transporter 2 (VMAT2). Retention of a tracer in vesicles is dependent on its structure – highly polar compounds like ¹¹C-EPI have long vesicular retention times, whereas more lipophilic compounds like ¹¹C-HED diffuse out of vesicles fairly quickly (5).

This approach to imaging sympathetic neurons has proven to be very effective, providing high quality cardiac images that roughly map the regional distribution of functional nerve terminals. However, a major limitation of the current generation of tracers is that their neuronal uptake rates, mediated by NET transport, are too rapid to allow accurate quantification of regional nerve densities. Rapid NET transport causes the net neuronal uptake of the tracers to be rate-limited by delivery from plasma to interstitium, rather than by NET-mediated neuronal uptake. Since delivery is governed primarily by perfusion, these are ‘flow-limited’ tracers. As a result, their myocardial kinetics cannot be successfully analyzed using standard kinetic analysis methodology. The inability to obtain quantitative measures using kinetic analysis methods forces the use of semi-quantitative measures of tracer retention as surrogate measures of nerve density, including the heart-to-mediastinum ratio (HMR) for MIBG and the retention index (RI) for ¹¹C-HED. However, due to the flow-limited uptake of the tracers, retention measures are insensitive to low-to-moderate levels of regional denervation, and decline only when nerve losses become fairly severe (6).

To overcome these limitations, our laboratory has been developing new tracers with more optimal kinetics for quantifying regional nerve density. Our goal was to develop a tracer with two specific kinetic characteristics. First, a slower NET transport rate was needed, making this the rate-limiting step in the neuronal uptake of the tracer. Second, efficient trapping inside storage vesicles was desired, leading to very long neuronal retention times (Fig. 1A). We hypothesized that a tracer with these properties could have its myocardial kinetics analyzed with a simple compartmental model in which the neurons act kinetically as a single irreversible compartment (Fig. 1B). In addition, these properties would allow the application of the Patlak graphical method for estimating the tracer’s net neuronal uptake rate (7).

FIGURE 1.

Comprehensive compartmental model of a sympathetic nerve radiotracer with optimal kinetic properties (A). Arrow thicknesses are drawn in approximate proportion to the magnitude of the rate constants. If the NET transport rate of the tracer (k₃) is less than the efflux rate from interstitium back into plasma (k₂), this prevents tissue uptake of tracer from being ‘flow-limited’. Furthermore, if the vesicular storage rate (k₅) is very rapid relative to efflux from neuronal axoplasm (k₄), and the tracer is effectively trapped in vesicles (k₆ = 0 or small), the two neuronal compartments kinetically behave as a single trapping compartment. This allows the use of the simplified compartmental model (B) for quantitative analysis of kinetic data from PET studies.

To this end, our studies of a series of ¹¹C-labeled phenethylguanidines identified N-¹¹C-guanyl-(−)-meta-octopamine (¹¹C-GMO, Fig. 2) as a promising tracer exhibiting the two targeted properties (8). The goal of the current study was to test our hypothesis that ¹¹C-GMO would exhibit myocardial kinetics that could be successfully analyzed with tracer kinetic techniques. We also evaluated the ability of quantitative measures from analyses of ¹¹C-GMO kinetics to track declines in available NET densities, induced pharmacologically with the potent NET inhibitor desipramine (DMI).

FIGURE 2.

Structure of N-¹¹C-guanyl-(−)-meta-octopamine (¹¹C-GMO). The carbon-11 label is incorporated into the guanidine group (*).

MATERIALS AND METHODS

Radiochemistry

¹¹C-GMO and ¹¹C-MIBG were synthesized as previously described (8). ¹¹C-HED was prepared using previously published methods (9). Specific activities for these compounds were 18.5–55.0 TBq/mmol. 2-¹⁸F-fluoro-2-deoxy-D-glucose (¹⁸F-FDG) was prepared as previously described (10).

Animal Care.

The care of all animals used in this study was done in accordance with the Animal Welfare Act and the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals (11). Animal protocols were approved by the University Committee on Use and Care of Animals (UCUCA) at the University of Michigan.

Isolated Rat Heart Studies

Kinetic studies of the neuronal uptake and retention of ¹¹C-GMO, ¹¹C-HED and ¹¹C-MIBG were performed with an isolated working rat heart system (8,12) under moderate workload conditions (10 cm H₂O preload, 100 cm H₂O afterload). For comparison, a study with ¹⁸F-FDG was performed under identical conditions except that the perfusate glucose concentration was changed from 5 to 10 mM. A 10 min constant infusion of tracer at very low concentrations was performed to measure neuronal uptake rates. The heart was then switched to a second perfusion circuit with normal perfusate for 120 min to study tracer clearance from sympathetic nerve terminals. Coincidence count rate data (cps/heart) from two opposing 5.1 × 5.1 cm cesium fluoride (CsF) detectors were corrected for random coincidences and normalized to the activity concentration in perfusate (cps/mL perfusate) and the heart’s wet mass (g wet/heart) to express heart uptake of tracer as an ‘apparent distribution volume’ (ADV; mL perfusate/g wet). Neuronal uptake rates K_up (mL perfusate/min/g wet) were determined as the linear slope the ADV data between 1 and 4 min of the constant infusion study. Clearance kinetics were fit to multiple exponential decay processes to characterize clearance rates. For the nerve tracers, 54 μM corticosterone was added to perfusate to block extraneuronal uptake (uptake-2) into myocytes (13), which competes with neuronal uptake (uptake-1) in rat hearts, but is absent in nonhuman primate and human hearts (6).

PET Imaging

PET imaging was performed using a Concorde Microsystems microPET P4 primate scanner (19 cm field of view, 7.8 cm axial extent, 1.75 mm intrinsic spatial resolution and 2.25% peak system sensitivity) (14). After the monkey was anesthetized, a percutaneous angiocather was placed in the saphenous vein of each leg (one for tracer injection, one for blood sampling). Heart rate (bpm), blood oxygen saturation levels (SpO₂) and body temperature were monitored continuously (SurgiVet V3404P). A transmission scan was acquired using a rotating ⁶⁸Ge/⁶⁸Ga rod source for attenuation corrections. Dynamic PET data were acquired in list-mode for 60 min after ¹¹C-GMO injection (24-44 MBq/kg). List-mode emission data were rebinned into a 24-frame dynamic sequence (12×10 s, 2×30 s, 2×60 s, 2×150 s, 2×300 s, 4×600 s). Rebinned emission data were corrected for attenuation and scatter, and transaxial images reconstructed using maximum a posteriori (MAP) reconstruction (15), an iterative method that accounts for the detector point spread function in the model of the system.

Radiometabolite Analyses

Before imaging, a blood sample (1.5-2.0 mL) was drawn and 3.7 MBq of ¹¹C-GMO was added. This was incubated at 37 °C for 60-70 min to determine tracer stability in blood. During the PET scan, four venous blood samples (1.5-2.0 mL) were drawn (t = 5, 15, 30 and 55 min) to assess radiometabolites in plasma and partitioning of ¹¹C-GMO between plasma and red blood cells (RBCs). Blood samples were centrifuged for 1 min at 12000 × g to separate plasma and RBCs. Plasma was deproteinized by adding perchloric acid (HClO₄; final concentration 0.4N) and centrifuging for 5 min at 12000 × g. The supernatant was neutralized with KOH (pH 7.0-7.5), filtered twice (0.22 μm filters; Millipore Millex/GS) and analyzed by HPLC (Synergi 10μ Hydro-RP column, 4.6 × 250 mm, 60 mM sodium phosphate buffer, pH 5.4, with 3% ethanol, flow rate 1.0 mL/min) and radiation detection (Ortec 905-4 NaI(T1) detector). The sample ‘spiked’ with ¹¹C-GMO was processed in the same way. Aliquots (0.1 mL) of whole blood, plasma, final supernatant and pellets were counted in a gamma counter. Count data (decay corrected) were used to determine the relative concentrations of ¹¹C-GMO in plasma and whole blood (C_p/C_wb). HPLC/radiation detection data (decay corrected) were processed for peak analysis (ACD/ChromProcessor v.10; ACD Inc., Toronto, Canada) to determine the percentage of activity associated with intact ¹¹C-GMO (f_intact). A mathematical function describing the time course of the metabolic breakdown of ¹¹C-GMO in plasma, f_intact(t), was obtained by nonlinear regression analysis (Prism 3.0, GraphPad Software, San Diego, CA).

Tracer Kinetic Analyses

Summed images of the final four PET frames were used to draw regions-of-interest (ROIs) on the myocardial wall and on the blood pool in the basal left ventricular chamber to extract time-activity curves for myocardial tissue C_t(t) and whole blood C_wb(t). The plasma concentration of intact ¹¹C-GMO vs. time, C_p(t), was estimated by multiplying C_wb(t) by the metabolic breakdown function, f_intact(t), and the mean ratio of activity concentrations in plasma and whole blood, C_p/C_wb. Thus, C_p(t) = C_wb(t)·f_intact(t)·[C_p/C_wb]. The plasma time-activity curve C_p(t) was used with the tissue time-activity curve C_t(t) for compartmental modeling. Using in-house analysis software, nonlinear regression analysis with the simplified compartmental model (Fig. 1B) provided estimates of the rate constants K₁ (mL/min/g), k₂ (min⁻¹), k₃ (min⁻¹) and a blood volume fraction BV (dimensionless). Rate constant estimates were used to calculate a ‘net uptake rate’ constant K_i (mL/min/g) = (K₁k₃)/(k₂ + k₃), which reflects the rate of ¹¹C-GMO accumulation into nerve terminals. Tissue and plasma kinetics of ¹¹C-GMO were also analyzed using Patlak graphical analysis (7). After construction of a Patlak plot, the last 9 points of the plot were analyzed with linear regression to determine the Patlak slope, K_p (mL/min/g). Under ideal conditions, the Patlak slope K_p is a direct measure of the ‘net uptake rate’ constant K_i, and thus for the model structure shown in Fig. 1B, is also equal to (K₁k₃)/(k₂ + k₃).

Control and Pharmacological Blocking Studies

In addition to control conditions (n = 4), NET blocking studies were performed with the NET inhibitor desipramine (DMI). The goal of these studies was to assess the ability of the quantitative parameters from kinetic analyses to track progressively lower cardiac NET densities induced pharmacologically by increasing DMI doses. All studies were performed in the same monkey to minimize biological variation between studies and to make initial assessments of the reproducibility of quantitative measures of NET density. The DMI dose (dissolved into 2.0 mL of sterile saline) was infused intravenously over 20 min using an infusion pump. ¹¹C-GMO was injected 10 min after the end of the DMI infusion. DMI doses used were: 0.010 mg/kg (n = 2), 0.0316 mg/kg (n = 2), 0.10 mg/kg (n = 2), 0.316 mg/kg (n = 1) and 1.0 mg/kg (n = 1). The measured kinetic parameters K_i from compartmental modeling and K_p values from Patlak analysis, as a function of DMI dose, were fit to a sigmoidal dose-response model with variable slope using nonlinear regression (Prism 3.0, GraphPad Software, San Diego, CA).

RESULTS

Isolated Rat Heart Studies

The neuronal uptake and retention kinetics of ¹¹C-GMO in the isolated rat heart are shown in Fig. 3A, in comparison to those of ¹¹C-MIBG and ¹¹C-HED. ¹¹C-GMO possesses a much slower neuronal uptake rate (K_up) than the other two tracers (Table 1), about 8 times slower than ¹¹C-HED and 12 times slower than ¹¹C-MIBG. ¹¹C-GMO also has a much longer neuronal retention time than ¹¹C-MIBG and ¹¹C-HED (Table 1). A slower neuronal uptake rate and a very long retention time are the two kinetic properties we hypothesized a nerve tracer would require for its kinetics to be analyzed successfully with tracer kinetic techniques. It is also interesting to compare the kinetics of ¹¹C-GMO with those of the well-established cardiac tracer ¹⁸F-FDG (10 mM glucose, no insulin) in this model (Fig. 3B). The uptake kinetics of the two tracers are almost identical, while ¹⁸F-FDG clears from myocardium with a major half-time of 4.6 h compared with a neuronal clearance half-time >200 h for ¹¹C-GMO (Table 1). Since coronary flow rates are more than 10 times physiological levels in this system, tracers diffusing from tissue spaces tend to be cleared from the isolated heart at faster rates than would be seen in vivo. Thus storage of ¹¹C-GMO inside vesicles is a very effective trapping mechanism leading to extremely long neuronal retention times.

FIGURE 3.

Kinetics of ¹¹C-MIBG, ¹¹C-HED and ¹¹C-GMO in isolated working rat hearts (A). In each case, an 10 min constant infusion of tracer was performed to measure neuronal uptake rates (K_up; mL/min/g wet), then the heart switched to normal heart perfusate to study efflux rates from neuronal spaces. For comparison, the isolated rat heart kinetics of ¹¹C-GMO are shown along with those of ¹⁸F-FDG (B). Note the different y-axis scales used in panels A and B.

Table 1.

Uptake rates (K_up) and major clearance half-times (T_1/2) in isolated rat hearts, as shown in Fig. 3.

Tracer	Kup (mL/min/g)	Major Clearance T1/2 (h)
11C-MIBG	3.65	2.1
11C-HED	2.35	1.1
11C-GMO	0.30	217
18F-FDG	0.31	4.6

¹¹C-GMO Metabolism

In monkey studies, ¹¹C-GMO was metabolized into one major and two minor metabolites, all of which were more polar than the parent compound (Fig. 4A). The half-time for the metabolic breakdown of ¹¹C-GMO averaged 16.0 ± 3.0 min (range 11.0–22.5 min). In most cases, the fraction of plasma activity associated with intact ¹¹C-GMO vs. time was fit to the following function (Fig. 4B):

$$f_{\text{intact}}(t)=Span\cdot \left(\frac{B^n}{t^n+B^n}\right)+\mathit{\text{Plateau}}$$ Eq. 1

FIGURE 4.

Reverse-phase HPLC/radiodetection analysis of radiometabolite formation in plasma (A). The fraction of activity associated with intact ¹¹C-GMO was determined for each sample, and the % intact vs. time data fit to a mathematical function to characterize the metabolic breakdown of ¹¹C-GMO (B).

For the 1.0 mg/kg DMI study, it was necessary to use an alternative function:

$$f_{\text{intact}}(t)=Span\cdot \left(\frac{U{e}^{-At}-A{e}^{-Ut}}{U-A}\right)+\mathit{\text{Plateau}}$$ Eq. 2

The fitted curve for f_intact(t) was used in preparing the ‘input function’ C_p(t) for kinetic analyses. HPLC analyses of blood samples ‘spiked’ with ¹¹C-GMO showed only the parent compound, indicating that it is stable in blood and plasma. Similar to ¹²³I-MIBG and ¹¹C-HED, ¹¹C-GMO is metabolically stable inside neurons. The guanidine group in the side chain of ¹¹C-GMO prevents metabolism by neuronal enzymes such as monoamine oxidase (MAO), although some phenethylguanidines are reversible MAO inhibitors (16). ¹¹C-GMO lacks the catechol structure of catecholamines like norepinephrine, so it is not metabolized by catechol-O-methyl-transferase (COMT).

¹¹C-GMO Partitioning in Blood

Analysis of ¹¹C-GMO concentrations in whole blood, plasma and red blood cells (RBCs) demonstrated that ¹¹C-GMO stays primarily in plasma with little uptake into RBCs. The ratio of plasma and whole blood activity concentrations, C_p(t)/C_wb(t), tended to be constant throughout the PET study. For all of the blood samples drawn during PET scanning, the average ratio was 1.47 ± 0.08. Similarly, for all ‘spiked’ blood samples, the mean ratio was 1.41 ± 0.07. DMI block of cardiac NET had no effect on the blood partitioning of ¹¹C-GMO. Blood partitioning data for each study was used in the preparation of ‘input functions’ for kinetic analyses.

Effects of DMI Infusion and ¹¹C-GMO Injection

Body temperature and SpO₂ levels were stable during all studies. For DMI levels at 0.10 mg/kg and below, DMI infusion caused no change in heart rate. However, for DMI doses of 0.316 mg/kg and 1.0 mg/kg, starting 3 min into the DMI infusion, heart rate steadily rose from a baseline of around 100 bpm up to 122-125 bpm at 1 min after DMI infusion. Heart rate then slowly declined 4-8 bpm over the next 60 min. Intravenous injection of ¹¹C-GMO had no effect on heart rate, with only a transient increase of 1-2 bpm for 1 min observed during the 0.316 mg/kg DMI study.

Imaging Properties

Representative transaxial PET images from a control study and progressively higher DMI block studies are shown in Fig. 5. The images clearly show a DMI dose-dependent decline in the cardiac retention of ¹¹C-GMO. In controls, heart-to-blood ratios of 3.01 ± 0.18 and heart-to-liver ratios of 0.89 ± 0.24 were seen in the final image (n = 4). If the relatively high liver uptake of ¹¹C-GMO also occurs in human subjects, this could be a minor drawback as the liver is often close to inferior or inferoseptal segments of the left ventricle, leading to some ‘spillover’ of counts from liver into these segments.

FIGURE 5.

Representative transaxial microPET images of cardiac ¹¹C-GMO retention. Shown are summed images (final four dynamic frames) for a control study (far left) and for blocking studies with progressively higher doses of the NET inhibitor desipramine (DMI).

Kinetic Analysis – Compartmental Modeling

Compartmental modeling of ¹¹C-GMO kinetics proved to be very robust, in the sense that the nonlinear regression algorithm converged in a few iterations to a single global minimum. Myocardial time-activity curves and corresponding compartmental model fits for a control study and three of the DMI blocking doses are presented in Fig. 6A-6D. Model parameter estimates for all studies are given in Table 2. Parameter estimates were fairly consistent within groups, but values of k₃, which ideally should reflect cardiac NET density, did not decrease with increasing DMI doses. However, the net uptake rate constant K_i calculated from the combination of estimated rate constants (K₁k₃)/(k₂ + k₃) did provide a quantitative measure that sensitively tracked reductions in available cardiac NET in a DMI dose-dependent manner (Table 2; Fig. 6E). Declines in K_i values (Y) vs. increasing DMI doses were well described by a sigmoidal dose-response model with variable Hill slope (n_H), where X = log [DMI dose]:

$$Y=\frac{Y_{\max }}{1+{10}^{[\log ({\mathrm{IC}}_{50})-X]\cdot n_{\mathrm{H}}}}$$ Eq. 3

FIGURE 6.

Compartmental modeling of ¹¹C-GMO kinetics using the simplified model shown in Fig. 1B. Myocardial ¹¹C-GMO kinetics (blue dots) and corresponding compartmental model fits (red lines) are shown for a control study (A) and three different DMI blocking doses (B-D). (E) Relationship between the net uptake rate constants K_i (mL/min/g) calculated from compartmental model parameter estimates and log [DMI dose (mg/kg)]. The decline in K_i values was well described by a sigmoidal dose response curve with variable Hill slope (n_H). Estimated parameter values for the dose-response curve fit are shown.

Table 2.

Results from compartmental modeling and Patlak analysis of ¹¹C-GMO kinetics.

	Compartmental Modeling					Patlak Analysis
Study	K1 (mL/min/g)	k2 (min−1)	k3 (min−1)	BV	Ki (mL/min/g)	Kp (mL/min/g)
Control - A	0.392	0.145	0.074	0.322	0.133	0.104
Control - B	0.345	0.123	0.082	0.283	0.138	0.111
Control - C	0.306	0.082	0.061	0.278	0.131	0.102
Control - D	0.282	0.117	0.087	0.252	0.120	0.099
0.01 mg/kg DMI - A	0.245	0.129	0.099	0.343	0.106	0.077
0.01 mg/kg DMI - B	0.259	0.138	0.089	0.356	0.101	0.073
0.0316 mg/kg DMI - A	0.271	0.255	0.114	0.361	0.084	0.062
0.0316 mg/kg DMI - B	0.256	0.188	0.094	0.326	0.085	0.064
0.1 mg/kg DMI - A	0.417	0.794	0.149	0.411	0.066	0.047
0.1 mg/kg DMI - B	0.217	0.240	0.113	0.370	0.069	0.051
0.316 mg/kg DMI	0.662	2.228	0.120	0.215	0.034	0.030
1.0 mg/kg DMI	0.102	0.435	0.074	0.592	0.015	0.017

‘Net uptake rate’ constants K_i were calculated from estimated parameter values as: (K₁k₃)/(k₂ + k₃).

For the K_i data, nonlinear regression analysis yielded parameter estimates of: IC₅₀ = 0.087 ± 0.012 mg/kg DMI, Hill slope n_H = –0.70 ± 0.07 and a maximum net uptake rate constant Y_max = 0.130 ± 0.003 mL/min/g (r² = 0.99). Since the Hill slope is less than –1.0, this indicates that the dose-response curve is more shallow than the standard one-site competition dose-response curve used frequently for in vitro competitive displacement assays. The observation that estimates of the individual model parameters K₁, k₂, k₃ and BV were variable and that k₃ estimates alone did not serve as the index of NET density was not unexpected, as a similar situation exists for analyses of ¹⁸F-FDG kinetics in assessments of cardiac glucose utilization (17).

Kinetic Analysis – Patlak Graphical Analysis

Patlak analysis of ¹¹C-GMO kinetics provided highly linear Patlak plots under all experimental conditions, with linear correlation coefficients r > 0.99 in all cases (Fig. 7A). Measured Patlak slopes K_p (mL/min/g) are shown in Table 2. Similar to the K_i values from compartmental modeling, the measured Patlak slopes K_p declined along a sigmoidal dose-response curve (Fig. 7B) with increasing doses of DMI: IC₅₀ = 0.068 ± 0.010 mg/kg DMI, Hill slope n_H = −0.54 ± 0.05 and a maximum Patlak slope Y_max = 0.104 ± 0.002 mL/min/g (r² = 0.99). There was a strong linear correlation between the measured Patlak slopes and the calculated K_i values from compartmental modeling: K_p = (0.757)K_i + 0.001, with linear correlation coefficient r = 0.99 (data not shown).

FIGURE 7.

Patlak plots of ¹¹C-GMO kinetics for all studies (A). Legend values are listed in order of highest Patlak slope to lowest, corresponding to values given in Table 2. (B) Relationship between Patlak slopes K_p (mL/min/g) and log [DMI dose (mg/kg)]. Reductions in measured K_p values vs. DMI dose were again well described by a sigmoidal dose response curve with variable Hill slope (n_H). Estimated parameter values for the dose-response curve are shown.

DISCUSSION

This study was designed to test if quantitative measurements derived from analyses of myocardial ¹¹C-GMO kinetics are capable of sensitively tracking cardiac sympathetic nerve densities over the full dynamic range seen in normal subjects and in patients with heart disease. The experimental approach used was to pharmacologically induce varying degrees of cardiac NET inhibition using the NET inhibitor desipramine (DMI). Our findings demonstrated that two measures of the ‘net uptake rate’ constant K_i – obtained either by calculation from compartmental modeling results or as the Patlak slope K_p from Patlak graphical analysis – each appear to serve as robust and reproducible measures of regional cardiac sympathetic nerve density. To the best of our knowledge, ¹¹C-GMO is the first sympathetic nerve radiotracer to possess myocardial kinetics that can be analyzed in a straightforward and robust manner using conventional kinetic analysis methods.

Isolated rat heart studies demonstrated that NET transports ¹¹C-GMO at a rate about 8 times slower than ¹¹C-HED (Fig. 3A). Similarly, NET transport assays using a rat C6 glioma cell line stably transfected with the cloned human NET (C6-hNET cells) showed that ¹¹C-GMO was transported at a rate about 9 times slower than ¹¹C-HED (18). We believe that this large reduction in the NET transport rate reduces the rate constant k₃ in the compartmental model to a magnitude that is less than k₂ (Fig 1B). When k₃ < k₂, the tracer’s net tissue uptake is not confounded by flow-limitation effects. As shown in Table 2, for all studies performed, k₃ was estimated to be less than k₂, which is consistent with the hypothesis that the net neuronal uptake of ¹¹C-GMO is not flow-limited. Thus we believe that the quantitative parameters K_i and K_p are likely to be sensitive measures of regional cardiac nerve density, capable of detecting mild-to-moderate nerve losses earlier in the progression of denervation than is currently possible with flow-limited tracers like ¹¹C-HED or ¹²³I-MIBG. The dose-response curves shown in Figs. 6E and 7B support this contention. DMI-induced declines in both K_i and K_p were well described by a sigmoidal dose-response model, with r² > 0.99 in each case. This means that more than 99% of the total variance in the net influx constants (K_i, K_p) is explained by variation in the DMI dose (as modeled by the dose-response function), demonstrating the high sensitivity of these measures.

Not surprisingly, there was a high correlation between K_i and K_p, which are essentially two measures of the same aggregate combination of the model rate constants, K₁k₃/(k₂ + k₃). In most cases, Patlak slopes K_p were a little lower than their corresponding K_i values. For Patlak analysis we did not make any corrections to the tissue kinetics for the presence of radioactivity in blood vessels or for spillover effects between blood pool and myocardium. This resulted in a modest downward bias in the Patlak slopes. The compartmental model’s ‘blood volume fraction’ (BV) parameter corrects for blood-borne radioactivity in tissue and for spillover effects (19). While K_p was lower than K_i for most studies, for the highest DMI concentration (1.0 mg/kg), K_p was higher than K_i (0.17 mL/min/g vs. 0.15 mL/min/g, respectively; Table 2). In this study, tissue activity levels were actually lower than blood activity levels late in the study, leading to a higher than average BV estimate, reflecting net spillover from blood into tissue. It appears that the compartmental modeling approach appropriately accounts for blood volume and spillover effects, and thus the calculated K_i constant should be more accurate than the Patlak slopes. That said, both measures are seen to provide an excellent index of NET transport rate.

We postulated that the main factor causing quantitative analyses of a tracer like ¹¹C-HED to fail was its rapid NET transport rate, resulting in flow-limited uptake. The present results demonstrate that the slower NET transport rate of ¹¹C-GMO, combined with its efficient trapping in storage vesicles, leads to myocardial kinetics that can be analyzed successfully. However, it is possible that another factor may contribute to successful analysis of ¹¹C-GMO kinetics – the tendency of the tracer to stay in plasma throughout the study, with very little trapping in RBCs. This allows intact ¹¹C-GMO molecules in plasma to accumulate slowly into nerve terminals during the entire PET study. ¹¹C-HED on the other hand tends to equilibrate with RBCs, reaching a C_p/C_wb ratio of 1.0 several minutes after tracer injection (20). If ¹¹C-HED molecules associated with RBCs are unavailable for extraction into tissue, preventing further accumulation into neurons, this may explain in part why cardiac ¹¹C-HED levels do not climb significantly after the rapid cardiac uptake seen early in a PET study (21). ¹¹C-HED’s uptake into RBCs may simply be due to passive diffusion, as it is more lipophilic than ¹¹C-GMO. However, the biogenic amines dopamine, norepinephrine and epinephrine are all actively transported into RBCs (22), so it is possible that this mechanism is also involved.

Measurement of Regional Norepinephrine Uptake Rates

In cardiac PET studies of glucose metabolism with ¹⁸F-FDG, it has been shown that regional metabolic rates of glucose utilization rMR_glu (μmol/min/g) can be calculated from net uptake rate constants K_i (or measured Patlak slopes K_p) (23). For such calculations, it is necessary to know the value of the ‘lumped constant’ (LC) in the heart (the correction factor which relates the observed ¹⁸F-FDG kinetics to those of glucose) as well as the subject’s plasma glucose concentration C_glu (μmol/mL). Then rMR_glu can be calculated as:

$${\text{rMR}}_{\text{glu}}(\mu \text{mol}/\text{min}/\mathrm{g}){=K}_{\text{i}}(\text{mL/min/g})·\frac{{\text{c}}_{\text{glu}}(\text{μmol/mL})}{\text{LC}}$$ Eq. 4

In a similar fashion, it may be possible to use regional K_i values from compartmental modeling of ¹¹C-GMO kinetics (or Patlak slopes K_p) to noninvasively estimate regional norepinephrine (NE) uptake rates (rUR_NE) in the heart, in units of pmol NE/min/g. This would require knowledge of the relative NET transport rates of norepinephrine and ¹¹C-GMO, here termed a ‘relative transport constant’ TC = UR_GMO/UR_NE. In addition, the plasma concentration of norepinephrine C_NE (pmol NE/mL) would need to be measured for the subject. Then rUR_NE could be calculated as:

$${\text{rUR}}_{\text{NE}}(\text{pmol}\mathrm{NE}/\text{min}/\mathrm{g}){=K}_{\text{i}}(\text{mL}/\text{min}/\mathrm{g})\cdot \frac{{\text{C}}_{\text{NE}}(\text{pmol}\mathrm{NE}/\text{mL})}{\text{TC}}$$ Eq. 5

As a starting point, TC can be approximated from isolated rat heart kinetics. In this system, ¹¹C-GMO has a NET-mediated uptake rate constant K_up equal to 0.30 mL/min/g (8), while the value for ³H-norepinephrine is 4.44 mL/min/g (24), so that TC = 0.30/4.44 = 0.068. Plasma norepinephrine levels C_NE for female rhesus macaques held in captivity have been reported to be approximately 1.2 pmol/mL (25). Using K_i = 0.130 mL/min/g for ¹¹C-GMO, an estimate of rUR_NE in the rhesus macaque heart is:

$${\text{rUR}}_{\text{NE}}(\text{pmol}\mathrm{NE}/\text{min}/\mathrm{g})\text{=}(0.130\text{mL}/\text{min}/\mathrm{g})·\frac{(1.2\text{pmol}/\text{mL})}{0.068}=2.3\text{pmol}/\text{min}/\mathrm{g}$$

Values for comparison are difficult to find, but Eisenhofer et al. estimated a whole-heart UR_NE = 819 pmol/min in human heart using ³H-norepinephrine spillover techniques (26). Taking a nominal mass of 300 g for a normal adult human heart, an estimated regional rUR_NE value would be: (819 pmol/min)/(300 g) = 2.7 pmol/min/g. These two estimates of rUR_NE agree well enough to suggest that this approach may provide a noninvasive method of measuring regional norepinephrine uptake rates in human subjects. Further work is needed to validate this approach. While such measurements might not find routine clinical use, they may find application in clinical studies of the effects of diseases on regional norepinephrine reuptake rates, or on the efficacies of drug therapies designed to improve or otherwise modulate norepinephrine reuptake rates.

CONCLUSION

¹¹C-GMO appears to be the first cardiac sympathetic nerve tracer possessing kinetics that can be analyzed in a straightforward manner to provide robust and sensitive quantitative measures of regional cardiac sympathetic nerve density. These quantitative measures will likely be able to detect mild-to-moderate sympathetic nerve losses that occur early in the course of cardiac denervation, which is not possible with existing tracers like ¹¹C-HED and ¹²³I-MIBG. In addition, ¹¹C-GMO may find application in monitoring the efficacy of novel drug therapies designed to halt or reverse cardiac denervation in diseases associated with enhanced risk of sudden cardiac death, such as diabetic autonomic neuropathy and heart failure. Finally, it may be possible to noninvasively quantify regional norepinephrine reuptake rates through extension of the methodology currently used for measuring glucose metabolic rates from parameters measured from ¹⁸F-FDG kinetics.