Introduction

The liver is central to the regulation of glucose and lipid metabolism, and is strongly implicated in the pathogenesis of diabetes [1]. The disease is commonly associated with higher circulating free fatty acid (FFA) levels, and increased FFA delivery to, and fat accumulation within, the liver have been shown to enhance gluconeogenesis, thus increasing hepatic glucose output and hyperglycaemia [2, 3, 4]. So far, only invasive measurements across the whole splanchnic bed have been available in humans, yielding estimates of net FFA exchange in visceral districts [5, 6]. With the exception of a few highly invasive experiments conducted during abdominal surgery in the 1970s [7], there has been no way of isolating the liver from the other splanchnic tissues in humans.

The purpose of this study was (a) to target liver FFA uptake in vivo by employing positron emission tomography (PET) in combination with the long-chain fatty acid analogue 14(R,S)-[18F]fluoro-6-thia-heptadecanoic acid ([18F]FTHA) in humans, (b) to evaluate the use of graphical analysis for the quantification of liver FFA influx rate constant and uptake over time, and (c) to explore the need for a metabolite-corrected [18F]FTHA blood input function and its impact on final data. Overnight fasted, healthy human subjects were included in the study. Results on skeletal muscle FFA metabolism in this population have been previously published [8] and were used here to evaluate their relationship with liver data.

Materials and methods

Subjects

Eight healthy men (age 48±3 years; BMI 25±3 kg m−2) with no family history of diabetes took part in the study. They had normal physical examination, with a blood pressure of 127±12/74±6 mmHg, normal electrocardiogram, normal exercise echocardiogram and normal routine blood tests. None of them was taking any medication. Written informed consent was obtained from all subjects, after the study had been approved by the Ethics Committee of Turku University Central Hospital and Kuopio University Hospital.

PET scanning

Studies were conducted after a 12- to 15-h overnight fast, using an ECAT 931-08/12 scanner (CTI Inc, Knoxville, Tenn.). [18F]FTHA was synthesised as previously described, leading to a radiochemical purity of >98% [8]. Two venous catheters were inserted, one into an antecubital vein for tracer injection and one into a heated vein (~70°C) for sampling of arterialised venous blood. After optimisation of the subject's position, transmission scans were obtained with a germanium-68 ring source to correct all subsequent emission data for tissue attenuation of gamma photons. Then, [18F]FTHA (171±5 MBq) was injected, and a 32-min dynamic scan was carried out to image the liver and the cardiac chamber, followed by a 15-min dynamic scan of the femoral region. Arterialised blood samples were collected throughout scanning to measure glucose, insulin, FFA and plasma [18F]FTHA metabolites as previously described [8].

Image processing

All sinograms were corrected for tissue attenuation, dead time and decay, and reconstructed using standard reconstruction algorithms. Final in-plane resolution was 9.5 mm full-width at half-maximum.

Large circular regions of interest (ROIs) for hepatic [18F]FTHA time-activity measurements were placed on two to four consecutive image planes in the right lobe of the liver. Such measurements were averaged to generate one tissue time-activity curve per patient. Smaller ROIs were drawn on two adjacent image planes in the left ventricular chamber of the heart for the measurement of radioactivity in arterial blood; special attention was paid to avoid contamination from surrounding myocardial tissue. Input functions were corrected for time delay by comparison with corresponding tissue time-activity curves, based on the method proposed by van den Hoff et al. [9] for studies with oxygen-15-labelled water. In our application, a two-tissue compartment model was used, and dispersion was excluded from the model.

Data analysis

Arterial and tissue time-activity curves were analysed graphically to derive the fractional uptake constant (K i), which is given by the slope of the linear fit of the data [10]. FFA uptake was computed as the product of K i and serum FFA levels. FFA uptake was calculated using both uncorrected and metabolite-corrected input curves. The non-metabolised fraction of [18F]FTHA was determined by high-performance liquid chromatography from nine blood samples, and the interpolated metabolite curve was used to correct the original data, as previously described [8]. Liver K i and FFA uptake were measured over different time intervals to evaluate their stability over time. Skeletal muscle FFA uptake was derived from the region of the thigh as previously described [8].

Statistical analysis

All data are presented as mean±SD. Differences in paired data were evaluated using Student's paired t test for repeated measurements.

Results

Plasma glucose and insulin levels were 5.0±0.4 mmol l−1 and 26±6 pmol l−1, respectively. Serum triglycerides were 0.59±0.16 mmol l−1.

Image quality was good, as exemplified in Fig. 1. Rapid accumulation of tracer in the liver was observed with time, leading to progressively higher tissue to blood radioactivity ratios (Fig. 2).

Fig. 1.
figure 1

Images of tracer distribution over time in a higher (top panel) and a lower plane (bottom panel) are shown from one study subject, demonstrating progressive tissue tracer accumulation and clear identification of the heart chamber and liver

Full size image
Fig. 2.
figure 2

Blood (▼) and tissue (■) time activity curves in one study subject

Full size image

Using metabolite-corrected input function curves, linear fit to the data was excellent (Fig. 3), with r values exceeding 0.99 in all subjects during each fitting time frame. Slightly higher r values were observed when the linear fit was started after 10 min from tracer injection (Table 1). Liver uptake results were minimally affected by the choice of the fitting interval (Fig. 4), and the spread around the population mean (coefficient of variation = SD/mean) remained equally small at each time interval, with a mean value of 31%±1% and a range of 1% (30–31%). The use of metabolite-uncorrected input functions significantly worsened the spread of data around the fitted line (Table 1) and led to a remarkable underestimation of liver FFA uptake at all time intervals (Fig. 4). The coefficient of variation across different time frames was 37%±8% with a range of 22% (29–51%).

Fig. 3.
figure 3

Linear fit to the data using graphical analysis and metabolite-corrected input function in one study subject, showing high concordance between measured and fitted results

Full size image
Table 1. Linear fit comparison using metabolite corrected (M) or original (O) input functions
Full size table
Fig. 4.
figure 4

Liver FFA uptake results in the study population (mean±SD) using metabolite-corrected (white bars) and uncorrected input curves (black bars), during progressive fitting intervals (starting time as indicated on the x axis, ending time 32 min). With corrected curves, results were consistent, showing a slight time-dependent decline (P<0.01); with uncorrected curves, higher variability and increasing underestimation of FFA uptake were consistently observed. *P<0.01 vs metabolite-corrected input function

Full size image

Individual results derived during the 10 to 32 min fitting time are given in Table 2. Values of liver [18F]FTHA K i and FFA uptake were 0.34±0.04 ml min−1 ml−1 and 0.20±0.06 µmol min−1 ml−1, respectively, and were ~50 times higher than those reported in skeletal muscle.

Table 2. Individual FFA uptake data in the liver and skeletal musclea
Full size table

Discussion

Our study provides the first non-invasive quantification of FFA uptake rate in the liver of humans in vivo. Our data are in full agreement with those previously measured in dogs [11], and by highly invasive procedures in humans [7]. In these studies, values of hepatic FFA extraction fraction of 0.22–0.31 in dogs and ~37% in humans were reported, which compare very well with the 0.27–0.40 range of K i values obtained here; extraction fraction and K i are related through blood flow, which approximates 1 ml min−1 ml−1 in the liver [12, 13]. We showed that each ml of liver took up an amount of FFAs which was around 50 times higher than that of 1 g of muscle. In a whole organism, given an average liver volume of 1.5 l and a muscle mass of 40% of body weight (=31±1 kg in our subjects), the liver contributed a total FFA uptake of ~302±34 µmol min−1 versus ~132±17 µmol min−1 for total skeletal muscle mass. This makes the liver by far the most important organ in fasting FFA turnover in the whole body, and highlights the magnitude of the FFA load that is continuously handled by the organ. Even small changes in its regulation deserve attention. A reduced uptake of FFA would result in higher circulating concentrations, which are strongly implicated in the development of insulin resistance and diabetes [2]. On the other hand, increased hepatic FFA uptake might promote fat deposition inside the organ, decrease glucose consumption and upregulate hepatic glucose production [3]. Studies addressing the occurrence of such phenomena in insulin-resistant patients are lacking.

We showed that [18F]FTHA-PET is a particularly well-suited tool for the study of FFA uptake by the liver. In the fasting state, FFAs entering the organ are mostly oxidised, and small amounts are stored in triglycerides for later release [6, 11]. The current evidence suggests that FTHA follows the same routes, with the advantage that, after transport into mitochondria and formation of two acetyl-CoA moieties, it cannot be further metabolised. The tracer would therefore be trapped intracellularly in proportion to its uptake, reflecting FFA storage and oxidation rates [14, 15, 16]. Based on this knowledge, we opted to use graphical analysis to quantify liver FFA uptake, a choice that was supported by a strong concordance between fitted and measured data. Together with a rapid and elevated tracer uptake, the high tissue to blood radioactivity ratio observed provided good quality images with negligible background noise. Results were consistent over time and inter-individually, with minimal improvement by fitting after the first 10 min from tracer injection. Our results underscore the need for metabolite correction of input data, as persistent underestimation and higher variability was observed when neglecting this factor, with growing bias of FFA uptake estimates as time progressed. Tissue metabolites were not measured at this stage. Though the demonstration of a best fit with metabolite-corrected input curves and the incapability of this tracer to progress through β-oxidation do not seem to speak in favour of plasma metabolite uptake in the liver, in vitro measurements are required to draw definitive conclusions. FFA uptake was calculated by multiplying the influx rate constant by peripheral serum FFA concentrations. Though the liver receives FFAs by an arterial and a portal venous route, the latter being inaccessible in humans, the assumption made here that FFA concentrations in these two blood pools are similar has been previously supported [7].

In conclusion, our data support the use of non-invasive quantification of hepatic FFA uptake in humans, showing the liver to be the leading organ in FFA handling. [18F]FTHA-PET appears to be a valuable tool for further investigation of hepatic FFA turnover in humans.