Brain metabolism of 13N-ammonia during acute hepatic encephalopathy in cirrhosis measured by positron emission tomography


  • Potential conflict of interest: Nothing to report.


Animal studies and results from 13N-ammonia positron emission tomography (PET) in patients with cirrhosis and minimal hepatic encephalopathy suggest that a disturbed brain ammonia metabolism plays a pivotal role in the pathogenesis of hepatic encephalopathy (HE). We studied brain ammonia kinetics in 8 patients with cirrhosis with an acute episode of clinically overt HE (I-IV), 7 patients with cirrhosis without HE, and 5 healthy subjects, using contemporary dynamic 13N-ammonia PET. Time courses were obtained of 13N-concentrations in cerebral cortex, basal ganglia, and cerebellum (PET-scans) as well as arterial 13N-ammonia, 13N-urea, and 13N-glutamine concentrations (blood samples) after 13N-ammonia injection. Regional 13N-ammonia kinetics was calculated by non-linear fitting of a physiological model of brain ammonia metabolism to the data. Mean permeability–surface area product of 13N-ammonia transfer across blood–brain barrier in cortex, PSBBB, was 0.21 mL blood/min/mL tissue in patients with HE, 0.31 in patients without HE, and 0.34 in healthy controls; similar differences were seen in basal ganglia and cerebellum. Metabolic trapping of blood 13N-ammonia in the brain showed neither regional, nor patient group differences. Mean net metabolic flux of ammonia from blood into intracellular glutamine in the cortex was 13.4 μmol/min/L tissue in patients with cirrhosis with HE, 7.4 in patients without HE, and 2.6 in healthy controls, significantly correlated to blood ammonia. In conclusion, increased cerebral trapping of ammonia in patients with cirrhosis with acute HE was primarily attributable to increased blood ammonia and to a minor extent to changed ammonia kinetics in the brain. (HEPATOLOGY 2005;43:42–50.)

The role of ammonia in the pathogenesis of hepatic encephalopathy (HE) and possible deleterious effects on brain energy metabolism and neurotransmission has been extensively studied.1–4 The effects of experimental porta-caval anastomosis with or without pharmacological manipulations of ammonia metabolism have been examined in rat brain,5–13 and the effects of excess ammonia have been tested in astrocyte cell cultures.14 Human studies of cerebral uptake and metabolism of ammonia included postmortem examinations of brain tissue from patients dying of HE15 and biochemical analysis of jugular vein blood samples.16–18 More recently, ammonia metabolism has been tested in positron emission tomography (PET) studies of monkeys,19 healthy humans,19-22 and patients with cirrhosis and minimal HE.20–23 The 13N-ammonia PET technique can, in conjunction with proper compartmental analysis, give quantitative estimates of kinetic parameters in awake subjects.

Results of early PET studies by Phelps et al.19 indicated the presence of regional variations in cerebral 13N-ammonia uptake in normal human subjects. Lockwood et al.20 subsequently emphasized that the metabolism of ammonia to glutamine in brain influences cerebral ammonia uptake and later calculated a global rate of cerebral 13N-ammonia metabolism in patients with cirrhosis with minimal HE and healthy controls.21, 22 Recently, Ahl et al.,23 using dynamic PET studies, found no significant difference of the permeability–surface area product of the transfer of 13N-ammonia across the blood–brain barrier (BBB) between patients with cirrhosis with minimal HE and patients with cirrhosis without HE. No PET studies have been published in patients with cirrhosis with clinically manifest HE. Nevertheless, the previously mentioned studies support a hypothesis that blood–brain transfer of ammonia and the subsequent intracellular metabolism of ammonia are both increased in patients with cirrhosis with HE. To address this issue, we performed dynamic 13N-ammonia PET studies in brain of patients with cirrhosis with an acute episode of overt HE, patients without HE, and healthy controls. To this end, we made use of the improved spatial and temporal resolution of contemporary PET technology. Kinetic parameters of transfer of 13N-ammonia (i.e.,13NH3 + 13NH4+) from blood to cells as well as intracellular metabolism of 13N-ammonia metabolism were estimated separately. Dynamic 13N-ammonia PET in conjunction with individual measurements of blood contents of 13N-ammonia and 13N-metabolites were employed for optimal kinetic modeling of the ammonia metabolism. In addition, regional cerebral blood flow was measured by dynamic 15O-water scans to obtain parameters of regional cerebral ammonia kinetics based on individual estimates of cerebral 13N-ammonia clearances and cerebral blood flow.


HE, hepatic encephalopathy; PET, positron emission tomography; BBB, blood brain barrier; 13N-ammonia, sum of 13N-ammonium ion (13NH4+) and 13N-ammonia free base (13NH3); CBF, regional cerebral blood flow; PS, permeability surface product; VOI, volume of interest.

Patients and Methods

Human Subjects.

Eight patients with cirrhosis and an acute episode of clinically overt type C HE, West Haven criteria grades I to IV, Glasgow coma scale score 1 to 12,24–26 7 patients with cirrhosis with no recent or previous clinical symptoms of HE, and 5 healthy individuals participated in the study (Table 1).

Table 1. Clinical Data
SubjectsLiver FunctionHE
 NM/FAge (y)GEC pt/refPro-thrombin (index)Albumin (μmol/L)Child-Pugh ClassArterial pCO2 (kPa)Arterial Standard HCO3 (mmol/L)Arterial pHArterial Ammonia (μmol/L)New Haven Grade
  1. NOTE. Patient GEC relative to that for a healthy subject of same age and body weight49; Prothrombin index = coagulation factors II + IV + X relative to mean control value. Child–Pugh class50 and HE grade24: number of patients.

  2. Abbreviations: N, number of subjects; M/F, number of males/females; GEC, galactose elimination capacity (mmol/min) pt/ref.

  3. *Continuous variables are given as mean (range).

Cirrhosis with acute HE85/360 (50–77)0.48 (.27–.61)0.39 (.27–.56)360 (274–491)A: 0 B: 1 C: 73.91 (3.71–4.27)23.7 (18.3–27.5)7.45 (7.38–7.49)73 (29–97)I: 1; II:3 III: 2; IV:2
Cirrhosis without HE74/354 (47–67)0.49 (.38–.59)0.41 (.27–.57)412 (272–449)A: 2 B: 0 C: 55.35 (4.27–6.62)27.0 (23.4–28.1)7.43 (7.38–7.48)50 (28–78)no HE
Healthy controls52/347 (41–56)1.0 (.85–1.15)1.0 (0.83–1.13)656 (602–726)5.45 (4.83–5.68)25.2 (24.0–27.1)7.42 (7.41–7.43)16 (9–20)

The diagnosis of cirrhosis was based on histology of liver biopsy specimens in each patient; all patients with cirrhosis had esophageal varices, indicating portosystemic vascular shunting. The cause was chronic alcoholism in 13 patients, hemochromatosis in 1, and cryptogenic in 1. Precipitating factors for the HE were bleeding from esophageal varices (n = 5) and infection (n = 3); in no cases were dehydration, diuretic, or other medications considered precipitating factors. All patients with cirrhosis had severely reduced metabolic liver function (Table 1); patients with HE had respiratory alkalosis that was only in part metabolically compensated (Table 1); blood ammonia was increased in patients with cirrhosis, especially in those with HE (Table 1). No patients were treated with branched-chain amino acids, but all HE patients were treated with lactulose. All patients were spontaneously normo-ventilating and hence none was sedated, intubated, or mechanically ventilated.


The study was performed according to the Helsinki II declaration and approved by the Ethics Committee of Aarhus County. Informed consent was obtained from each subject, except those patients with HE grades III to IV, for whom informed consent was obtained from a next-of-kin. Healthy controls were recruited through an announcement in a local newspaper. The total radioactive dose received by the individuals was, on average, 2.2 mSv. No complications of the procedures occurred in any subject.

PET Scanning.

Scanning was performed by using a Siemens ECAT EXACT HR PET tomograph (CTI/Siemens, Knoxville, TN). Tracers were produced at the PET center by applying standard techniques and commercially available systems for isotope generation (General Electric, Uppsala, Sweden). The subject was positioned with the brain within the 15-cm field of view of the scanner and the head in a custom-made holder to prevent movements during and between studies. Dynamic emission recordings were made after 2 successive intravenous injections of 500 MBq 15O-water and 2 successive injections of 500 MBq 13N-ammonia. Eight patients with cirrhosis had only one 13N-ammonia scan because of technical problems. 15O-water was administered as a bolus injection over 4 seconds and 13N-ammonia as a bolus injection over 15 seconds to ensure an adequate number of blood radioactivity measurements in the initial dynamic phase for precise calculation of kinetic parameters.

The dynamic 15O-water study consisted of 21 time frames: 10 × 3 s, 3 × 10 s, and 8 × 15 s (3 minutes). The 13N-ammonia study consisted of 27 time frames: 6 × 10 s, 8 × 15 s, 2 × 30 s, 3 × 60 s, 4 × 120 s, 2 × 150 s, and 2 × 300 s (30 minutes). Data were recorded as mean values in each time frame, corrected for attenuation based on an initial 15-minute transmission scan using external Ge 68/64 rotating sources, and corrected for radioactive decay to start of the scanning. Reconstruction was performed by a 2D back-projection algorithm, resulting in 3-dimensional images consisting of 128 × 128 × 47 voxels of 2.0 × 2.0 × 3.1 mm3. Central spatial resolution was 4.5 mm full-width at half-maximum.

Specific brain regions were identified in images of the mean radioactivity concentrations from the dynamic 13N-ammonia emission recordings after the first 13N-ammonia injection (Fig. 1). As seen in Fig. 1, brain regions were quite well defined on this image presentation, and by simultaneous inspection of an average magnetic resonance imaging image based on 7 patients with cirrhosis and 5 healthy controls (Keiding, unpublished data, July 2005), using the Talairach and Tournoux model27 as a template, regions of interest were drawn in adjacent transaxial slices. The regions of interest were summed to volumes of interest (VOIs) for the right and left frontal cortex, temporal cortex, parietal cortex, occipital cortex, caudate nuclei, thalamus, putamen, and cerebellar lobes. No significant left–right differences were seen in the kinetic parameters within the cortical regions, the basal ganglia, or the cerebellar regions (data not shown). Hence, further analyses were performed for 3 combined VOIs: cortex, basal ganglia, and cerebellum. The VOIs were subsequently placed on images of the mean radioactivity concentrations of the second 13N-ammonia scan and of the 15O-water scans. Finally, the VOIs were transferred to the respective dynamic images, and time courses of tissue radioactivity concentrations were extracted (Fig. 2).

Figure 1.

PET images of transaxial (top) and sagittal (bottom) 3-mm slices showing mean brain 13N-radioactivity concentrations 0-30 minutes after intravenous injection of 500 MBq 13N-ammonia in a patients with cirrhosis with an acute episode of HE. PET, positron emission tomography; HE, hepatic encephalopathy.

Figure 2.

Time course of 13N-radioactivity concentrations during 30 minutes in cerebral cortex and arterial blood of 13N-ammonia, 13N-urea, and 13N-glutamine after intravenous injection of 500 MBq 13N-ammonia in a healthy individual. Measured data points are connected by straight lines. Model fit to the data is shown as a curve.

Blood Sampling.

Before scanning, a short catheter (Artflon, Becton Dickinson, Swindon, UK) was placed percutaneously in a radial artery under local anesthesia. During the 15O-water scans, arterial blood was sampled by an automated blood sampling system28 cross-calibrated with the tomograph. Radioactivity concentrations were measured every 0.5 seconds and corrected for external delay and dispersion in the sampling catheter and for radioactive decay to start of the scan.

During the 13N-ammonia scans, a series of 24 arterial blood samples of 1 mL each were collected manually at the following intervals: 12 × 5 s, 3 × 10 s, 1 × 30 s, 1 × 60 s, 2 × 120 s, 1 × 180 s, and 4 × 300 s. Total blood 13N-radioactivity concentrations were measured by using a well counter (Packard Instruments Co., Meriden, CT), cross-calibrated with the tomograph and corrected for radioactive decay to start of the scan.

Additional blood samples were collected at 1, 2, 3, 5, 7, 10, 15, 20, 25, and 30 minutes after the 13N-ammonia injection for measurements of 13N-metabolites by solid phase extraction according to Rosenspire et al.,29 validated by radio–high-pressure liquid chromatography of blood samples at 5 and 15 minutes.30 Detection of 13N-ammonia and 13N-urea were verified by co-injection of 13N-labeled ammonia and by ultraviolet detection of non-radioactive urea. Based on former studies identifying 13N-glutamine by an enzymatic technique, we attributed the single remaining high-pressure liquid chromatography fraction to 13N-glutamine. Other radiochemical fractions were insignificant or constituted less than 5% of recovered radioactivity. A double-exponential curve was fitted to the time course of the 13N-ammonia fraction, and mono-exponential curves were fitted to the 13N-urea and 13N-glutamine fractions. These fittings were used to estimate continuous individual time–activity curves for 13N-ammonia, 13N-urea, and 13N-glutamine (Fig. 2).

Arterial blood ammonia,31 pH, standard HCO3, and pCO2 (ABL, Radiometer, Copenhagen, Denmark) were measured 3 times during the 13N-ammonia recordings. None of the measurements changed significantly in the course of the scanning period, and mean values were used (Table 1).

Kinetic Analysis.

Regional cerebral blood flow (CBF) was calculated according to a standard 15O-water one-tissue compartment model.32 Clearance of 15O-water from blood into brain parenchyma, K1 (mL blood/min/mL tissue), and the rate constant of back-flux from brain parenchyma to blood, k2 (per minute), were calculated by non-linear fitting of the model to the measured time courses of 15O-radioactivity concentration in tissue, and using the measured time course of arterial blood 15O-radioactivity concentration as input function.32 Regional CBF (mL blood/min/mL tissue), was calculated as CBF = K1/0.9, where 0.9 corrects for the extraction fraction of 15O-water in the brain.33

The experimental 13N-ammonia data were analyzed by the two–tissue compartment model of 13N-ammonia uptake and metabolism illustrated in Fig. 3, characterized by 13N-ammonia clearance from blood to cell across the BBB, K1 (mL blood/min/mL tissue), back-flux of 13N-ammonia by rate constant, k2 (per minute), and conversion of intracellular 13N-ammonia to 13N-glutamine by rate constant, k3 (per minute). 13N-urea formed in the liver was assumed to enter and leave brain by diffusion characterized by K*1 and k*2 equal to kinetic terms for 15O-water, assuming identical kinetics for the 2 tracers. The kinetic compartment for arterial blood 13N-ammonia and 13N-metabolites was fixed at 0.01 mL/mL in accordance with preliminary fitting (data not shown). Kinetic parameters were calculated by non-linear fitting of the model (Fig. 3) to the measured time courses of brain tissue 13N-radioactivity concentrations, using time courses of blood 13N-ammonia, 13N-urea, and 13N-glutamine as input functions for each subject and each of the 3 brain regions (see Fig. 2).

Figure 3.

Compartment model of 13N-ammonia metabolism in brain: 13N-ammonia transfer across the blood-brain barrier is characterized by clearance K1 (mL blood/min/mL tissue) and back-flux rate constant k2 (per minute). 13N-urea (formed in the liver) enters and leaves the brain cells characterized by K1* (mL blood/min/mL tissue) and k2* (per minute). Blood 13N-glutamine is formed in extracerebral tissues (including liver) and distributes in brain vascular volume.

We use the concept of flow-independent permeability–surface area product (PS)34, 35 to characterize (1) net transfer of 13N-ammonia from blood into brain parenchyma across the BBB, PSBBB (mL blood/min/mL tissue) and (2) net metabolic conversion of blood 13N-ammonia into intracellular 13N-glutamine, PSmet (mL blood/min/mL tissue):

PSBBB was calculated according to the Kety-Renkin-Crone relation as

equation image(1)

where K1/CBF equals the extraction fraction of the initial distribution of 13N-ammonia.

Net metabolic clearance of blood 13N-ammonia into intracellular 13N-glutamine, Kmet (mL blood/min/mL tissue), was calculated as the asymptote of the Gjedde-Patlak linearization35, 36 of the model and equals

equation image(2)

PSmet of the metabolic conversion of blood 13N-ammonia into intracellular 13N-glutamine was calculated as

equation image(3)

where Kmet/CBF is the extraction fraction of steady state ammonia metabolism.

Metabolic flux of ammonia molecules into glutamine, Fluxmet (μmol ammonia/min/L tissue) was calculated as

equation image(4)

where A is the arterial blood ammonia concentration (see Table 1).

Correlation between kinetic parameters and their identifiability were examined by sensitivity analysis.37 This showed that K1 and PSBBB were mainly determined from the early data when 13N-ammonia distributes across the BBB, and Kmet and PSmet were mainly determined from late data when enzymatic conversion of 13N-ammonia into 13N-glutamine dominates the PET signal. Mathematical independence of the kinetic parameters was shown by the fact that no sensitivity function was proportional to or a reciprocal of another.


Results are given as mean ± SEM. Calculation of the kinetic parameters from the 3 brain regions in individual subjects were based on the same input curve and thus cannot be assumed to be statistically independent. Accordingly, the Newman-Keuls test was used for comparisons between groups of subjects. In the case of significance, group-wise comparisons of regional parameters were performed by Wilcoxon's test for unpaired data. P < .05 was considered statistically significant.


Duplicate measurements of the regional cerebral blood flow (CBF) in each subject deviated less than 5%, and individual mean values were used for further kinetic calculations. CBF in cortex was on average approximately two thirds of the values in the basal ganglia and cerebellum in each group of subjects and approximately 10% lower in the patients with cirrhosis with HE than in the healthy controls (Table 2).

Table 2. Regional Cerebral Blood Flow and 13N-Ammonia Kinetic Parameters in Patients With Cirrhosis With and Without an Acute Episode of HE and in Healthy Controls
  CBFK1PSBBBKmetPSmetMetabolic flux
mL blood/min/mL tissuemL blood/min/mL tissuemL blood/min/mL tissuemL blood/min/mL tissuemL blood/min/mL tissueμmol ammonia/min/L tissue
  • NOTE. Values are given as mean ± standard error of the mean. The * and ** indicate that the mean is statistically significantly different from the corresponding regional value in healthy control group:

  • *

    P < .05

  • **

    P < .01. CBF, cerebral blood flow, is determined from dynamic 15O-water PET. Dynamic 13N-ammonia PET estimates: K1, initial clearance of 13N-ammonia across blood–brain barrier (BBB) from blood to cells; PSBBB, blood-brain permeability for 13N-ammonia; Kmet, net metabolic clearance of 13N-ammonia in blood from intracellular 13N-metabolites (mainly 13N-glutamine). Metabolic flux of ammonia from blood into brain tissue is calculated as Kmet x arterial ammonia concentration.

Cirrhosis with HE (n = 8)Cortex0.41 ± 0.040.29 ± 0.040.21 ± 0.02*0.15 ± 0.010.20 ± 0.0113.4 ± 0.9**
Basal ganglia0.60 ± 0.050.36 ± 0.050.28 ± 0.03*0.21 ± 0.010.27 ± 0.0118.2 ± 1.2**
 Cerebellum0.56 ± 0.060.38 ± 0.060.20 ± 0.02*0.19 ± 0.010.24 ± 0.0116.7 ± 1.3**
Cirrhosis without HE (n = 7)Cortex0.44 ± 0.030.32 ± 0.040.31 ± 0.03*0.20 ± 0.03*0.23 ± 0.047.4 ± 1.4**
Basal ganglia0.59 ± 0.040.36 ± 0.020.37 ± 0.070.27 ± 0.04*0.32 ± 0.0410.6 ± 2.0**
 Cerebellum0.58 ± 0.040.32 ± 0.030.27 ± 0.040.24 ± 0.030.27 ± 0.038.8 ± 1.6**
Healthy controls (n = 5)Cortex0.47 ± 0.030.29 ± 0.030.34 ± 0.030.16 ± 0.010.20 ± 0.012.6 ± 0.3
Basal ganglia0.62 ± 0.050.31 ± 0.010.44 ± 0.030.21 ± 0.010.26 ± 0.023.4 ± 0.3
 Cerebellum0.64 ± 0.040.30 ± 0.020.34 ± 0.020.19 ± 0.010.23 ± 0.013.1 ± 0.4

Duplicate measurements of clearance K1 of the BBB transfer of 13N-ammonia in 12 subjects deviated less than 13%, and individual mean values were used. There was significant difference neither between cerebral regions nor between groups of subjects (Table 2). The efflux rate constant for ammonia, k2 (mean, approximately 0.1/min) and the rate constant for the glutamine synthetase reaction, k3 (mean, approximately 0.1/min) varied considerably, but all estimates were positive and without systematic differences between groups of individuals or brain regions (data not shown). Efflux of 13N-glutamine from brain to blood was not detectable within the current 30-minutes study period, as showed by a linear asymptote of the Gjedde-Patlak regression in each case.

For PSBBB, duplicate determinations in 12 subjects deviated on average 17% (range, 1%-38%), and individual mean values were used. Rank order for mean values was cirrhosis with HE < cirrhosis without HE < healthy controls in each of the 3 brain regions (see Table 2). For each region, values in the group of patients with HE were approximately two thirds of those in the healthy control group (Table 2). There was a tendency to higher values in the basal ganglia than in the other regions in each group of subjects, but the difference did not reach statistical significance.

For PSmet, duplicate determinations in 12 subjects deviated on average 6% (range, 1%-24%), and individual mean values were used. Neither significant differences of the metabolic clearance Kmet, nor of the PSmet between groups of individuals or brain regions were seen, although both values tended to be lower in cerebral cortex than in the other regions (see Table 2).

Flux of non-radioactive blood ammonia into intracellular glutamine, Fluxmet, (equation 4) showed no significant regional variations, although the rank order was cortex < cerebellum < basal ganglia within each groups of subjects (Table 2). Mean values were remarkably higher in the 2 cirrhotic groups with cirrhosis than in the healthy control group, especially in the HE group (Table 2). Figure 4 shows a linear relation between ammonia flux and endogenous blood ammonia for each of the 3 brain regions. The slope for the cortex relation tended to be lower than that for the 2 other regions in accordance with the tendency for lower cortical Kmet values.

Figure 4.

Metabolic flux ammonia from blood into intracellular glutamine, Fluxmet, in relation to arterial blood ammonia for cerebral regions as indicated for patients with cirrhosis with HE (closed triangle), patients with cirrhosis without HE (circle), and healthy controls (open triangle). Straight lines show linear regressions through 0.0 with the slopes (i.e., average Kmet) indicated. HE, hepatic encephalopathy.


Brain ammonia metabolism constitutes 2 successive steps: (1) transfer across the BBB and (2) enzymatic conversion into glutamine in the astrocyte (Fig. 3). The permeability–surface area product PSBBB characterizes transfer of ammonia from blood to brain parenchyma across the BBB. There may be intracellular and extracellular compartments within the brain, but in the modeling they are united into a single tissue compartment. PSBBB was reduced in each of the 3 brain regions in patients with cirrhosis, especially in those with HE, compared with healthy controls (Table 2). The physiological parameter PSmet characterizes the combined transfer of ammonia from blood to brain parenchyma and the subsequent enzymatic conversion to glutamine; it was not significantly different between the 3 groups of individuals. As is discussed later, these results are in contrast to prevailing views that PS-values for 13N-ammonia are increased in HE.4, 20–22 Current methods, however, constitute a significant improvement through the use of the improved spatial and temporal resolution of contemporary PET technology, the use of selective measurements of labeled ammonia and its metabolites in blood to characterize the input function, and the application of physiological modeling for examination of complex metabolic processes in vivo.

The compartment model of 13N-ammonia distribution and metabolism illustrated in Fig. 3 is a simplified depiction of the true complexity of brain ammonia metabolism.4 This simplification is required by the inability of PET recordings to distinguish multiple metabolic compartments in the brain. We first explored a model that ignored blood–brain transfer of 13N-urea and cerebral distribution of 13N-glutamine and used only blood 13N-ammonia as the input; however, fitting of this model to the PET data was clearly biased. Separate accounting for blood 13N-metabolites improved the goodness of fitting considerably as evaluated by residuals of the fitting and the Akaike score.38 This finding is probably related to the rapid depletion of 13N-ammonia in blood and rapid 13N-metabolite formation (Fig. 2).

The variable, but positive k2 estimate in each brain region is consistent with a minor back-flux of 13N-ammonia, as must be the case for simple diffusion across the BBB.4, 39, 40 The rate constant for the glutamine synthetase reaction, k3, similarly varied but was positive in each case, and tended to be higher in the HE group than in the other groups. There was no detectable efflux of 13N-glutamine from brain to blood within the current 30-minute study period, which is in agreement with previous observations.5, 41, 42 Accordingly, brain 13N-concentration (Fig. 2) may be ascribed initially to passage of the blood tracer bolus and blood-to-brain transfer of 13N-ammonia and after the first few minutes to accumulation of 13N-glutamine, in agreement with findings in 13N-ammonia PET pig liver studies.43

There was a tendency to lower cerebral blood flow rates in the cirrhosis groups than in the control group (Table 2) in each of the 3 regions, but this did not reach statistical significance. Studies have shown CBF to reduce with aging44 and the current findings may be related to the somewhat higher mean age of the cirrhosis groups than that of the control group (Table 1). Regional variations similar to the current findings were found in 2 other 15O-water PET studies of healthy controls and patients with minimal HE.23, 45 The reasons for the variations are not known but may be related to variations in the distribution of carbohydrate45 and ammonia metabolism.20, 23

The reduction of the BBB permeability PSBBB for 13N-ammonia in HE patients to approximately two thirds of control values in cerebral cortex, basal ganglia, and cerebellum is a novel finding of the current study. Mean cortex PSBBB in the 4 patients with HE grades III to IV was 0.15 mL/min/mL tissue and that in patients with HE grades I to II 0.26 mL/min/mL tissue. No earlier study has reported measurements of BBB ammonia transfer in patients with cirrhosis with clinically overt HE, and the current results indicate that the more severe the HE grade, the lower the PSBBB. Ahl et al.23 found no significant differences of PSBBB of 13N-ammonia in 5 patients with cirrhosis with minimal HE and 3 patients with cirrhosis without HE; the average was 0.24 mL/min/mL tissue. The magnitude of this estimate is lower than in the current patients with cirrhosis without HE, approximately 0.31 mL/min/mL tissue (Table 2). However, we used individually measured corrections of 13N-metabolites in the input whereas Ahl et al.23 used a common 13N-metabolite correction taken from a separate study of healthy controls,29 a methodological difference that most likely explains the discrepancy between the results of the 2 studies. Regional variations, with a tendency to higher PSBBB values in the basal ganglia (Table 2), were found in both studies.

Transfer of ammonia across the BBB is generally thought to be mediated by simple diffusion of NH3 and specific NH4+ transport proteins.4 In that case, the slight alkalosis seen in our patients might tend to favor the NH3 transfer and thus increase the PSBBB. However, the opposite was actually observed, indicating that a pH effect cannot explain our findings. The reduced PSBBB in the patients with cirrhosis with HE might be linked to inflammatory factors46 or increased diffusion distance through an increased extracellular volume in patients with cirrhosis with low-grade cerebral edema.47 It also could be related to other abnormalities such as altered membrane properties attributable to long-term exposure to elevated blood ammonia in analogy with the decreased membrane fluidity observed in patients with chronic high alcohol intake.48 Finally, the activity of specific transport proteins for NH4+ in the BBB4 may be altered because of high ammonia concentrations.

The PSmet, the permeability–surface product of the metabolic conversion of blood 13N-ammonia into intracellular 13N-glutamine, tended to be higher in the basal ganglia than in the 2 other brain regions, but the difference did not reach statistical significance (Table 2). Distinguishing between kinetic parameters of the BBB (permeability PSBBB related to initial transfer of ammonia across the BBB) and metabolism (PSmet related to steady-state metabolic conversion of ammonia to glutamine in the cells) is important. Lockwood et al.,20–22 in 1979 published pioneer work of brain ammonia metabolism using 13N-ammonia scans and in 1984 and 1991 estimated a PS value that corresponds to PSmet: they multiplied the amount of tracer in the brain (static PET of the brain in the interval 10 to 25 minutes after tracer injection) relative to the amount of tracer in the body (scan from head to upper leg) with a clearance value estimated from the first 10 minutes after tracer injection of arterial time course of non-metabolized ammonia. Thus, they calculated the fraction of the total body clearance that on average was retained in the brain during the scan and then calculated a PS value according to equation 3. They did not use initial PET recordings within the very first minutes after tracer injection and thus did not estimate initial kinetic parameters such as PSBBB as in the current work and in that of Ahl et al.23 The current dynamic scanning sequences in conjunction with compartmental analysis use the better temporal and spatial resolution of contemporary PET technology and also specified the separate inputs of blood 13N-ammonia and 13N-metabolites, thus yielding results based on a superior physiological model. We therefore conclude that regional PSmet is not changed significantly in patients with cirrhosis with or without HE compared with healthy controls (Table 2).

The finding of increased flux of blood ammonia into intracellular glutamine, Fluxmet (Table 2 and Fig. 4), in patients with cirrhosis compared with healthy controls is similar to that found by Lockwood et al.20 The magnitude of Fluxmet was significantly correlated to blood ammonia (Fig. 4). The regional differences (Fig. 4) indicate that the basal ganglia and cerebellum might be more sensitive than cortex to changes in blood ammonia.

In conclusion, contemporary dynamic PET technique showed decreased BBB permeability for ammonia, PSBBB, in patients with cirrhosis with HE to approximately two thirds of control values. Secondly, PSmet of the cerebral metabolic trapping of ammonia as glutamine was unchanged in patients with cirrhosis, whereas the flux of blood ammonia into brain tissue was markedly increased in both patients with cirrhosis with and without HE, depending primarily on arterial blood ammonia. Thus, elevated blood ammonia in cirrhosis rather than changed ammonia kinetics determines brain uptake and possible neurotoxic effects of ammonia.


The authors thank Paul Cumming for pivotal reading of the manuscript.