FT

PLT

*R¶

Bfc(ptopu(batptPudbcm.bcarc(cuacccs

KL

Ti

CLIN

ICA

L–LIVER

,PA

NCREA

S,A

ND

BILIA

RY

TRA

CT

GASTROENTEROLOGY 2010;139:846–856

atty Acid Metabolism in the Liver, Measured by Positron Emissionomography, Is Increased in Obese Individuals

ATRICIA IOZZO,*,‡ MARCO BUCCI,*,‡ ANNE ROIVAINEN,* KJELL NÅGREN,* MIKKO J. JÄRVISALO,* JAN KISS,*,§

ETIZIA GUIDUCCI,‡ BARBARA FIELDING,� ALEXANDRU G. NAUM,* RONALD BORRA,* KIRSI VIRTANEN,*IMO SAVUNEN,§ PIERO A. SALVADORI,‡ ELE FERRANNINI,¶ JUHANI KNUUTI,* and PIRJO NUUTILA*

Turku PET Centre, and Department of Medicine, §Department of Surgery, University of Turku, Turku, Finland; ‡Institute of Clinical Physiology, PET Lab, Nationalesearch Council, Pisa, Italy; �Oxford Centre for Diabetes, Endocrinology, and Metabolism, University of Oxford, Churchill Hospital, Oxford, United Kingdom;

Department of Internal Medicine, University of Pisa School of Medicine, Pisa, Italy

bilcttpltm

eootstdogoso

ahftpptbomf

pr

ACKGROUND & AIMS: Hepatic lipotoxicity resultsrom and contributes to obesity-related disorders. It is ahallenge to study human metabolism of fatty acidsFAs) in the liver. We combined 11C-palmitate imaging byositron emission tomography (PET) with compartmen-al modeling to determine rates of hepatic FA uptake,xidation, and storage, as well as triglyceride release inigs and human beings. METHODS: Anesthetized pigsnderwent 11C-palmitate PET imaging during fasting

n � 3) or euglycemic hyperinsulinemia (n � 3). Meta-olic products of FAs were measured in arterial, portal,nd hepatic venous blood. The imaging methodologyhen was tested in 15 human subjects (8 obese subjects);lasma 11C-palmitate kinetic analyses were used to quan-ify systemic and visceral lipolysis. RESULTS: In pigs,ET-derived and corresponding measured FA fluxes (FAptake, esterification, and triglyceride FA release) did notiffer and were correlated with each other. In humaneings, obese subjects had increased hepatic FA oxidationompared with controls (mean � standard error of theean, 0.16 � 0.01 vs 0.08 � 0.01 �mol/min/mL; P �

0007); FA uptake and esterification rates did not differetween obese subjects and controls. Liver FA oxidationorrelated with plasma insulin levels (r � 0.61, P � .016),dipose tissue (r � 0.58, P � .024), and systemic insulinesistance (r � 0.62, P � .015). Hepatic FA esterificationorrelated with the systemic release of FA into plasmar � 0.71, P � .003). CONCLUSIONS: PET imagingan be used to measure FA metabolism in the liver. Bysing this technology, we found that obese individu-ls have increased hepatic oxidation of FA, in theontext of adipose tissue insulin resistance, and in-reased FA flux from visceral fat. FA flux from vis-eral fat is proportional with the mass of the corre-ponding depot.

eywords: Insulin Resistance; Lipolysis; Oxidative Stress;iver Steatosis.

he liver plays a central role as both a target and acause in obesity-related disorders. Hepatic lipotoxic-

ty is strongly implicated in this association, and may

ecome responsible for the majority of liver transplantsn future decades.1 Although hepatic abnormalities inipid metabolism have a benign course in the majority ofases, it is currently not possible to predict progressionoward organ disease. Some subjects may be committedo liver damage from the beginning of the disease,2

rompting the need to elucidate the sequence of eventseading to liver lipotoxicity and develop noninvasive toolso identify early risk individuals and devise effective treat-

ent.Excess accumulation of triglycerides in the liver is

xpanding at an epidemic rate, involving approximatelyne third of the adult population.3 Although the amountf hepatic fat is a marker of increased exposure of hepa-ocytes to potentially toxic fatty acids (FAs),4 recent datauggest that it has a protective role in sequestering FA,hus reducing their free intracellular levels.5,6 Instead, theamage of fat-laden hepatocytes results from reactivexygen species, causing hepatic inflammation and fibro-enesis.7,8 Therefore, the assessment of FA oxidation isne key element in the characterization of the high-riskituation because this process is the source of reactivexygen species in obesity-related hepatic lipotoxicity.7,9

The flux of FAs from white adipose tissue to the livernd its hormonal control are important determinants ofepatic FA exposure. The liver receives a dual FA supply,

rom the systemic circulation and from the portal vein;he latter drains the visceral compartment, which is ex-anded in obesity, and active in releasing FA.10 Insulinromotes nutrient storage both in adipose tissue and inhe liver; via its antilipolytic effect, it supports the FA-uffering capacity of triglyceride depots. Loss of controlf this process because of insulin resistance or deficiencyay lead to enhanced FA oxidation and free radical

ormation.

Abbreviations used in this paper: BMI, body mass index; CT, com-uterized tomography; FA, fatty acid; PET, positron emission tomog-aphy; TG, triglyceride.

© 2010 by the AGA Institute0016-5085/$36.00

doi:10.1053/j.gastro.2010.05.039

novcbFbvF

nPvtmtit

t

(tpoops

bi6ftct1

ttma

elPw

sw

ofcmabbotn(str

adpwAuf

lswpvobti

T

AS

D

BFGIHTHFH

F

NHh

CLI

NIC

AL–

LIV

ER,

PA

NCREA

S,A

ND

BIL

IARY

TRA

CT

September 2010 HEPATIC FATTY ACID METABOLISM 847

Direct assessment of liver metabolism by arterial-ve-ous differences is impractical in human beings becausef its invasiveness and the inaccessibility of the portalein. Positron emission tomography (PET) is a tool ofhoice for noninvasive molecular imaging, and it haseen used in combination with 11C-palmitate to quantifyA oxidative metabolism in the myocardium andrain,11–13 but not in the liver. The technique also pro-ides the plasma kinetics of labeled palmitate to describeA release from adipose tissue into the circulation.14

In the present work, we combined portal, hepatic ve-ous, and arterial catheterization with liver 11C-palmitateET imaging and compartmental modeling in pigs toalidate the use of this imaging technique in the quan-ification of liver FA metabolism. We then applied the

ethod to human beings, documenting the early activa-ion of hepatic FA oxidation in obese vs lean individualsn relation to systemic and adipose tissue insulin resis-ance and visceral fat lipolysis.

Materials and MethodsFor an extended description, see the Supplemen-

ary Materials and Methods section.

Animal StudiesAnesthetized pigs were studied during fasting

n � 3) or euglycemic hyperinsulinemia (n � 3). Cathe-ers were placed in the jugular vein, carotid artery, andortal, hepatic, and femoral veins, for the administrationf glucose, insulin, and 11C-palmitate, and for samplingf blood. Doppler flow-probes were placed around theortal vein. Vital signs were monitored throughout thetudy (Supplementary Figure 1).

PET imaging. A transmission scan was followedy 11C-palmitate injection and 50 minutes of dynamic

maging (20 frames, 8 � 15, 2 � 30, 2 � 120, 1 � 180,� 300, 1 � 600 s). Blood was drawn simultaneously

rom the carotid artery and portal vein at each imagingime frame to measure the total plasma radioactivity foromputation of the input function, and the carotid ar-ery and hepatic and portal veins for the measurement of1C-labeled FA, triglyceride, phospholipid, CO2, and wa-er-soluble pools, as described in the Supplementary Ma-erials and Methods section. At the end of each experi-

ent, animals were killed and the liver volume wasssessed.

Image processing. Large circular regions of inter-st were placed on consecutive image planes in the rightobe of the liver for hepatic radioactivity measurements.lasma radioactivity levels in the artery and portal veinere combined into one input function.15

Human StudiesThe evaluation in human beings included 15

tudies, 9 of which were performed ad hoc to image the

hole liver and visualize the portal vein carefully, as part m

f the input function needed to calculate FA metabolismrom imaging data. These 9 studies were first used toonfirm relationships with circulating markers of liveretabolism, and to validate the extension of the current

pproach to studies lacking portal vein information, toe used in the remaining 6 human studies. The latter hadeen performed previously to evaluate myocardial metab-lism11; thus, images included only the upper portion ofhe liver. Human subjects were divided into obese andonobese based on a body mass index (BMI) cut-off level

30 kg/m2); the one patient with full-blown metabolicyndrome but a BMI less than 30 kg/m2 was included inhe obese group (n � 8). Subjects’ characteristics areeported in Table 1.

PET imaging. Computerized tomography (CT)ttenuation and dynamic PET images of the upper ab-ominal area, lasting 50 minutes and including 20 sam-ling frames, as described in the Animal Studies section,ere performed after administration of 11C-palmitate.rterialized venous blood was withdrawn every 10 min-tes to discount the image-derived input function for the

raction of metabolized radiopharmaceutical.Image processing. Image reconstruction and de-

ineation of regions of interest was performed as de-cribed in the Animal Studies section. In 9 studies, inhich CT and image quality allowed identification of theortal vein, small regions of interest were drawn in thisessel and in the left ventricular cavity of the heart, tobtain radioactivity concentrations in portal and arteriallood, respectively, to be converted into plasma activityhrough the hematocrit. Because the dual-input functions not always resolvable from human images, we used a

able 1. Characteristics of Study Group

Controlsubjects(n � 7)

Obesesubjects(n � 8) P value

ge, y 49 � 1 55 � 3 .10ystolic bloodpressure, mm Hg

119 � 4 138 � 9 .07

iastolic bloodpressure, mm Hg

72 � 3 88 � 5 .018

MI, kg/m2 26 � 1 32 � 1 .004at mass, kg 24 � 2 34 � 2 .006lucose level, mmol/L 4.8 � 0.1 5.3 � 0.1 .032

nsulin level, mU/L 4.1 � 0.3 8.5 � 1.6 .024bA1c level, % 5.4 � 0.1 5.7 � 0.1 .08G level, mmol/L 0.74 � 0.25 1.70 � 0.37 .06DL level, mmol/L 1.43 � 0.11 1.53 � 0.17 .62ree FA level, mmol/L 0.51 � 0.04 0.63 � 0.05 .09OMA index, insulin �glucose/22.5

0.9 � 0.1 2.0 � 0.4 .02

at insulin resistance,insulin � free FA

2.2 � 0.2 5.7 � 8.4 .04

OTE. Data are given as mean � standard error of the mean.bA1, glycosylated hemoglobin; HDL, high-density lipoprotein; HOMA,omeostatic model assessment.

athematic model to derive average kinetics parameters

tfis(f

srma

oM

nitdttcp1

dcm(t

cip2owrFrFvum

mitecp

tlpasam1

baIccludbspw

tsaife

mvGRnt

trmatiuatwpmF

CLIN

ICA

L–LIVER

,PA

NCREA

S,A

ND

BILIA

RY

TRA

CT

848 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

hat would permit its estimation from the arterial inputunction, as shown in Supplementary Figure 2. Thisnput function was applied in 6 control individuals. Sub-equent data analysis was performed with either the dualmeasured and estimated) or the single arterial inputunction for comparisons.

Magnetic resonance studies. In the 9 humanubjects enrolled prospectively for this study, magneticesonance imaging and spectroscopy were performed to

easure intra-abdominal fat masses and liver fat content,s previously described.16

Sample AnalysesAssays of 11C-labeled CO2, water, and lipid metab-

lites are described in the Supplementary Materials andethods section.

Kinetic AnalysesCompartmental modeling of liver 11C-palmitate ki-

etics. Compartmental modeling describes each processnvolved in FA handling by the liver, including inwardransport, release of labeled 11C-products owing to oxi-ation of palmitate, incorporation of 11C-palmitate intoriglyceride, and the subsequent partial release of labeledriglycerides in the circulation. The model (Figure 1A)onsists of 3 tissue compartments, representing free 11C-almitate, 11C-palmitate bound in complex lipids, and

1C-oxidative breakdown products. The exchange of ra-ioactivity across compartments is described by 5 rateonstant terms, K1– k5. The differential equations of theodel and derivation of substrate steady-state flux rates

�mol/mL/min) are described in the Supplementary Ma-erials and Methods section.

Lipolysis and fatty acid clearance. The metaboliclearance rate of the tracer was estimated by dividing thenjected dose by the integral (from zero to infinity) of thelasma time-activity curve (input function obtained from0 image time points, as described earlier). Plasma curvesf 11C-palmitate in arterial and portal venous plasmaere used for the evaluation of the FA clearance rate in

espective circulatory beds, and multiplied by arterial freeA levels to obtain corresponding rates of appearance,eflecting adipose tissue lipolysis. The difference betweenA appearance in the 2 vessels was assumed to representisceral fat lipolysis, draining directly into the liver; these of arterial free FA values leads to a minimum esti-ate of visceral fat lipolysis.

Validation PointsModel fit. The goodness of fit of the model-esti-

ated vs the measured liver tissue time-activity curve wasnspected and quantified by the sum of square computa-ions, adjusted to the sample size and number of param-ters (Akaike17 information criterion), and the coeffi-ients of variation of each estimated rate constant

arameter. c

PET-derived vs measured FA fluxes. In animals,he hepatic output of 11C-oxidative products and 11C-abeled triglyceride (TG) was derived from PET data as aroduct of the estimated 11C-concentration in the lipidnd oxidative compartments (Figure 1C) and the corre-ponding outflow rate constants. The value at 40 minutesfter tracer injection was used for comparison against theeasured net balance of 11C-labeled water, 11C-CO2, and

1C-TG across the liver, computed by multiplying hepaticlood flow and the difference between flow-weightedrterial-portal and venous radioactivity concentrations.n addition, hepatic oxidative metabolism by PET wasompared with lipid oxidation, as obtained by indirectalorimetry equations applied to organ exchange of un-abeled oxygen and carbon dioxide.18 Finally, total FAptake rates obtained by mathematic modeling of PETata were compared with those measured through thealance of free and bound FA across the liver duringteady-state, which were computed by multiplying he-atic blood flow and the difference between flow-eighted arterial-portal and venous substrate levels.

Use of a single arterial input function and predic-ion of a dual-input function. In human beings, use of aingle arterial or a dual-input function derived fromrterial concentrations through compartmental model-ng vs the originally measured dual arterial-portal inputunction to estimate liver FA metabolic fate wasvaluated.

Statistical AnalysesResults are given as mean � standard error of the

ean. Comparison between model-derived and measuredariables were performed by the Student paired t test.roup differences were evaluated by analysis of variance.egression analyses were performed by standard tech-iques. A P value of less than .05 was considered statis-ically significant.

ResultsAnimal StudiesImage quality and model fit. Hepatic 11C-palmi-

ate images were of excellent quality, showing progressiveetention of radioactivity. Direct blood sampling in ani-

als showed that 11C-palmitate levels were higher inrterial than portal plasma only during the peak phase,hat is, in the first minute after tracer injection, converg-ng to similar values after approximately 3 minutes (Fig-re 1B). Therefore, in the human studies image-derivedrterial and portal venous measurements of 11C-palmi-ate concentrations were combined over peak values,hereas arterial determinations were used for later timeoints. Representative examples of model-predicted vseasured liver tissue time-activity curves are given in

igure 1C, also showing the simulated time-activity con-

entrations in each tissue compartment. The goodness of fit

w1a

Fp(ctw

wrsmcadFl

Fpvrki

CLI

NIC

AL–

LIV

ER,

PA

NCREA

S,A

ND

BIL

IARY

TRA

CT

September 2010 HEPATIC FATTY ACID METABOLISM 849

as evaluated as weighted sum of squares,17 which averaged.77 � 0.70. Model parameters did not show any significantutocorrelation (Supplementary Table 1).

PET-derived vs measured FA fluxes. As shown inigure 2, the model-predicted outflow of FA oxidativeroducts (kBq/min/mL of organ) and FA oxidative rate�mol/min/mL) indicated no significant difference asompared with respectively measured values. Correla-ions of estimated and measured labeled water products

igure 1. (A) Configuration of the model used to derive the rate constalasma into the tissue and between the cytoplasm, and the lipid and oxidenous concentrations of 11C-palmitate over time showing that the cuight). (C) The model fit to the measured tissue 11C-palmitate time-activiinetics with excellent approximation; the tracer radioactivity in each mod

n one study animal.

as significant (r � 0.91, P � .01), and a strong tendency T

as shown between estimated and measured FA oxidativeates (r � 0.63, P � .18), which fell short of statisticalignificance, likely owing to group size. In addition, the

easured rate of release of 11C-oxidative products wasorrelated significantly with both the FA oxidation ratend the FA concentrations in the oxidative (mitochon-rial) pool obtained from the model (Supplementaryigure 3). Similar correspondences were observed in re-

ation to TG metabolism and total hepatic FA uptake.

1–k5) describing the fractional inward/backward movement of FA frompool compartments. (B) A representative example of arterial and portalonverge after the first sampling minutes (expanded time scale on theve shows that the mathematic derivation predicts the measured tracermpartment also is shown, as derived from the estimated rate constants

nts (Kative

rves cty curel co

he model predicted vs measured labeled-TG release and

lasoswcF

sblpdwmctoTssde

h

tdawev5r

9er�mtelc1

0i

Ftos

F(T(

CLIN

ICA

L–LIVER

,PA

NCREA

S,A

ND

BILIA

RY

TRA

CT

850 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

iver FA uptake were similar between methods (Figure 2),nd the estimated vs measured uptake rate showed atrong tendency toward a correlation (r � 0.75, P � .15)f borderline significance because of a limited sampleize. Circulating TG levels were correlated significantlyith both model-predicted 11C-TG release rates and FA

oncentrations in the hepatic TG pool (Supplementaryigure 3).

Effects of insulin. The use of FA by the liver is aummation process, resulting from the combined contri-ution of FA delivery to the organ, that is, plasma FA

evels, and the intrinsic tissue extraction and fractionalartition of the substrate between oxidative and nonoxi-ative pathways. The model rate constants and the in-ard clearance rate represent the intrinsic tissue deter-inants because they are independent of circulating FA

oncentrations. Our animal experiments documented in-rinsic effects of insulin (Figure 3) to suppress the effluxf labeled TG from the liver, thereby likely promotingGFA retention in the storage pool. The corresponding

ubstrate fluxes, lumping together the intrinsic liver re-ponse and the extrinsic availability of circulating FA,ocumented a suppression of liver FA uptake, oxidation,sterification, and output.

Human StudiesImage quality and model fit. Images obtained in

igure 2. Comparison showing that the model predicted hepatic up-ake and oxidative or nonoxidative metabolism of FA is similar to valuesbtained by direct measurement of arterial-portal vs hepatic venousubstrate and products differences.

uman studies (Figure 4A) allowed extraction of portal i

ime activity curves. In the human studies, in which theual-input function was measured directly from the im-ges, the goodness of fit, as weighted sum of squares,17

as 0.78 � 0.40 (NS vs animal studies). The accuracy ofstimation of each parameter, that is, the coefficient ofariation, was 6% � 1%, 18% � 3%, 16% � 4%, and 19% �% for the inward, oxidative, esterification, and outputate constants, respectively.

PET-derived fluxes vs measured indicators. Thehuman studies in which the dual-input function was

xtracted directly from the PET-CT images were used aseference in the current validation. Circulating levels of-hydroxybutyrate, which is an indirect but organ-specificarker of liver FA oxidation, were associated positively with

he model-predicted oxidative rate (r � 0.76, P � .03), afterxcluding one subject in whom �-hydroxybutyrate was be-ow the detection level. Similar to findings in animals, cir-ulating TG levels in these subjects were correlated with1C-palmitate esterification and 11C-TG release rates (r �.80, P � .018). Whole-body lipid oxidation derived from

ndirect calorimetry assessments was associated with the

igure 3. The effects of insulin on the hepatic FA model rate constantstop) document a direct suppression of the fractional release of labeledG (magnified graph), whereas liver FA uptake and metabolic fluxes

bottom) are suppressed by indirect inhibition of circulating FA levels by

nsulin. Values are given as mean � standard error of the mean.

pe

tamocm

Co(umwcd

Foaot

CLI

NIC

AL–

LIV

ER,

PA

NCREA

S,A

ND

BIL

IARY

TRA

CT

September 2010 HEPATIC FATTY ACID METABOLISM 851

roportions of FA entering hepatic oxidation (positive) andsterification (negative).

Use of a single arterial input function and predic-ion of a dual-input function. The use of a single arterial vs

dual arterial-portal input function to estimate liver FAetabolic fate was associated with a 46% underestimation

f K1 (P � .004), similar oxidative and 11C-TG output rateonstants (P � NS), and an approximate 64% overesti-

igure 4. (A) Representative 2-dimensional and 3-dimensional PET-Cf the portal vein. (B) The arterial input function (left) is used to estimatend the estimated is compared with the measured curve on the right. (Cbtained with the original vs estimated dual- vs single-arterial input funcimation in the parameters derived from the third approach.

ation of the TG synthetic rate constant (P � .022). w

onsequently, rates of FA uptake (�46%; P � .003),xidative metabolism (�56%; P � .008), esterification�27%; P � .004), and release (�28%; P � .005) were allnderestimated using the single- vs dual-input functionodel (Figure 4C). Despite absolute differences, theeighted sum of squares (0.70 � 0.28),17 and coeffi-

ients of variation of individual parameters were notifferent, and proportional partition between path-

ges of hepatic 11C-palmitate distribution, allowing optimal visualizational-input function with the approach shown in Supplementary Figure 2,comparison between FA flux rates (mean � standard error of the mean), showing good correspondence between the first 2, and an underes-

T imaa du

) Thetions

ays was preserved.

dpwpatip

ruacfliii

Fhp( visce

CLIN

ICA

L–LIVER

,PA

NCREA

S,A

ND

BILIA

RY

TRA

CT

852 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

Instead, the use of the arterial input function to pre-ict the dual-input function by mathematic modelingrovided parameters describing the dual-input functionith excellent approximation (Figure 4B). The estimatedarameters did not correlate with the amount of intra-bdominal fat (Supplementary Table 2). These parame-ers were averaged and applied again to each arterialnput function, showing that the use of fixed-average

igure 5. (A) Metabolic flux rates of FA in the liver in obese and controlepatic FA oxidation in the former group, leading to the relative shift in faercentages on the right and rate constants on the left. (C) Lipolytic rate

left); the difference between the 2 measurements (right) represents the

arameters did not affect the model fit (Akaike17 crite- r

ion, 0.87 � 0.36) or coefficients of variation of individ-al parameters (P � NS as compared with the individu-lly measured dual-input function), or the ultimatealculation of liver FA metabolic rate constants anduxes, as given in Figure 4C. These parameters were used

n the remaining 6 subjects, lacking access to portal veinmages. The liver model parameters in the final analysisn human beings did not show any significant autocor-

cts (mean � standard error of the mean), showing a 2-fold increase inf the oxidative vs esterification pathway, as shown in panel B, given asdetermined from 11C-palmitate arterial and portal venous plasma levelsral contribution.

subjevor o

s, as

elation (Supplementary Table 1).

tootlswFaastdh

i

a(tfiecftowsa

ttabttsaminf

oipostptesetesvculwopaosefF

Fcacla

CLI

NIC

AL–

LIV

ER,

PA

NCREA

S,A

ND

BIL

IARY

TRA

CT

September 2010 HEPATIC FATTY ACID METABOLISM 853

Effects of obesity. In control subjects, FA oxida-ion and esterification rates accounted for 40% and 60%f liver FA uptake, respectively (Figure 5B). Hepatic FAxidation rates were on average 100% greater in obesehan control individuals (Figure 5A); thus, the intracel-ular partition between oxidation and esterification washifted significantly in favor of the former. No differencesere noted between men and women. The rate of systemicA appearance (ie, lipolysis) was not different between leannd obese individuals. However, the difference between FAppearing in the portal vs the systemic circulation wasignificant in obese subjects, and less pronounced in con-rol subjects (Figure 5C, left panel), resulting in a ten-ency (just short of statistical significance) toward aigher visceral fat FA contribution in the former group.Hepatic FA oxidation was associated with circulating

igure 6. (A) Regression analyses in 15 human subjects show signifi-ant positive relationships between hepatic FA oxidative metabolismnd indexes of insulin resistance, and (B) between hepatic FA esterifi-ation and extrahepatic lipolysis. (C) The visceral FA contribution to the

iver appears dependent on the mass of the corresponding fat depot,nd not on that of total body fat.

nsulin levels (r � 0.61, P � .016), homeostatic model C

ssessment index, and adipose tissue insulin resistanceFigure 6A). Systemic and portal FA appearance (lipoly-ic) rates were the strongest predictors of hepatic esteri-cation (Figure 6B) and release rate of FA in TGs. Thestimated visceral fat FA contribution to the liver wasorrelated with the intra-abdominal, but not total body.at mass (Figure 6C), supporting the current interpreta-ion of this variable. Liver FA uptake (ie, the sum ofxidative and nonoxidative components) was correlatedith homeostatic model assessment and fat insulin re-

istance indexes (P � .05) and systemic and portal FAppearance rates (P � .02).

DiscussionWe used 11C-palmitate PET imaging to quantify

he fate of FAs in the liver. In human beings, liver me-abolism generally is extrapolated from splanchnic bal-nce studies, which are invasive and cannot distinguishetween the relative contributions of gut, visceral adiposeissue, and liver. This limitation is especially important inhe metabolism of FA relative to other substrates becauseimultaneous FA uptake and release occur continuouslynd to variable extents in these organs. The presentethodology offers the important advantages of estimat-

ng the intracellular fate of FAs between the oxidative andonoxidative pathways, and the contribution of visceral

at to FA flowing to the liver.In the validation study, we opted for a limited number

f animals, which were studied in depth by parallel PETmaging and gold standard methodology, and we com-ared parameter estimation by compartmental modelingf liver 11C-palmitate kinetics with conventional flux as-essments obtained during fasting and insulin stimula-ion. The swine model was chosen for its organ size,ermitting the use of a human scanner, thus allowingest image quality and analysis tools for an immediatexportation into human beings, and easier access to theignificant blood sampling required in this study. FAsterification was significant in pigs, despite the notionhat these animals are not prone to develop steatosis. Anxtensive protocol was used to attain simultaneous mea-urements from the 2 perfusing and 1 draining liveressels, together with organ blood flow and volume, toompute steady-state influx-efflux differences. The totalptake of free and bound fatty acids, and the release of

abeled 11C-TG or 11C-oxidative products, were comparedith the corresponding estimates by PET. Because liverxidative metabolism includes generation of ketones byartial FA oxidation, and recycling of CO2 via productionnd release of urea and glucose, we assumed that thergan arterial-portal vs venous difference of 11C–water-oluble compounds and of 11CO2 encompasses all thearlier-described products and represents total hepaticatty acid oxidation, as previously performed by others.19

urthermore, we measured the exchange of oxygen and

O2 across the liver to calculate lipid oxidation rates by

uinflpcdsogt

aapwmhifgWuolatcosmsfmatttoWaeotbi1

apwip

pvtTtfpswapetiih

mepbsea

oFtfcogpoftbsccphonm2pmaaiaotv

CLIN

ICA

L–LIVER

,PA

NCREA

S,A

ND

BILIA

RY

TRA

CT

854 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

se of indirect calorimetry equations. The quality of PETmages indicated that the liver is FA avid, as confirmed byumeric results, showing that approximately 40% of FAsowing to the liver are extracted by the organ. Modelingarameters were robustly identifiable, as reflected by theiroefficients of variation and lack of autocorrelation. Theirectly measured and the PET-derived variables showedimilar and correlated results, and the imaging method-logy showed an increased sensitivity in discriminatingroup differences with sufficient statistical power despitehe limited sample size.

The liver receives dual blood supply from the hepaticrtery and portal vein, and the latter is not directlyccessible in human beings. We were able to extract theortal time activity curve from the images in the subjectsho were studied with PET-CT. However, this approachay be impractical if anatomic CT images are lacking in

uman PET scanners or if the lower portion of the livers not within the imaging field of view, and it remains farrom the resolution capabilities of small animal tomo-raphs, which are used increasingly in metabolic studies.e first explored the bias associated with the alternative

se of the single arterial input function in the modelingf liver 11C-palmitate data, and showed that this method

eads to an underestimation of hepatic substrate rates,ffecting more severely the earlier (uptake, oxidation)han the later (esterification, release) intracellular pro-esses. The metabolic pathways maintained some degreef reciprocal proportionality, suggesting that the methodtill may be acceptable for group comparisons. To obtain

ore accurate absolute estimates, we evaluated the pos-ibility of deriving kinetic parameters describing the dualrom the arterial input function, by using a compartment

odel that accounts for the delay and dispersion ofrterial tracer concentrations in the passage of bloodhrough the visceral compartments upstream to the por-al vein. Our results show an optimal correspondence ofhe measured and the estimated (either with individualr group-averaged rate constants) dual-input functions.ithin the range of intra-abdominal fat mass values

vailable in this study, covering 4-fold differences, thestimated parameters did not correlate with the amountf fat. However, we cannot rule out that the accuracy ofhe model for estimating the dual input may be affectedy the amount of visceral fat in more severely obese

ndividuals, who may have higher uptake and release of1C-palmitate by visceral adipose tissue.

Physiopathologic Findings: Initial Elements inthe Cascade of Liver FA LipotoxicityThe current dataset documents that insulin exerts

direct effect on TGFA output from the liver, thusromoting lipid storage. This finding identifies the path-ay of insulin action surmised in previous human stud-

es, in which the acute inhibition of TG-rich lipoprotein

roduction by insulin implicated an FA-independent he- c

atic mechanism.20,21 The current result extends to the inivo situation the evidence shown in vitro that the addi-ion of insulin to hepatocyte cultures reduced the rate ofG secretion, and this was the only process in TG me-

abolism that was affected by the hormone in a directashion.22 Accordingly, the pronounced reduction in he-atic FA uptake, oxidation, and esterification rates ob-erved in our study in response to insulin stimulationas mediated entirely by the hormonal suppression ofdipose tissue lipolysis. This dual action of insulin ap-ears to rule out a role of this hormone in the directnzymatic control of liver FA esterification and oxida-ion. This finding needs to be confirmed in human be-ngs, and implies that liver metabolism may not be stud-ed in isolation from the concomitant control of theepatic-adipose axis.Our obese group presented prodromal features of theetabolic syndrome, including overweight, hyperinsulin-

mia, and mild insulin resistance, slight increases inlasma glucose and TG levels, and a mild increase inlood pressure. Circulating free FAs were not increasedignificantly, presumably because of the hyperinsulin-mia, thus minimizing any interference of substrate massction on measured hepatic FA metabolism.

Our key finding was a 100% increase in hepatic FAxidation in obese individuals, in the face of a preservedA uptake and esterification rate. We previously showedhat impaired glucose tolerance is associated with a de-ect in hepatic FA uptake, and we hypothesized the oc-urrence of substrate competition (ie, increased glucosexidation owing to hyperglycemia).23 The current studyroups had normal glucose tolerance, and the fastinglasma glucose levels, although marginally higher in thebese subjects, were below the threshold of impairedasting glucose. The effect of obesity could not be inves-igated in the previous study23 because those patients hadeen matched purposely for BMI. The current datasethowed that, in the absence of hyperglycemia, obesity isharacterized by preserved hepatic FA uptake and in-reased FA oxidation. Circulating insulin levels and adi-ose tissue insulin resistance were main correlates ofepatic FA oxidation, and the hepatic exposure to FAwing to the contribution of visceral fat was more pro-ounced in obese than control individuals by approxi-ately 40 �mol/min, corresponding to approximately

0% of systemic lipolysis, in line with the estimates re-orted by others.10 Our findings show by direct measure-ent that the primary hepatic response to obesity is the

ctivation of FA oxidation as a consequence of systemicnd adipose tissue insulin resistance and visceral adipos-ty.7 This observation corroborates previous evidence innimals, showing increased capacity for mitochondrialxidation and greater activities of citrate synthase andotal carnitine palmitoyltransferase in genetically obeses lean rodents,24 and supports the concept that a defi-

ient insulin action may reduce the affinity of carnitine

pi

todttilsoFcbscto

acwlosFsfovhpfitagsnectltrm

ifootctr

ab

aG1

1

1

1

1

1

1

1

1

1

CLI

NIC

AL–

LIV

ER,

PA

NCREA

S,A

ND

BIL

IARY

TRA

CT

September 2010 HEPATIC FATTY ACID METABOLISM 855

almitoyltransferase for malonyl-CoA, thus limiting thenhibitory effect of the latter on the former.25

Beyond its protective action to remove free FA fromhe intracellular space, FA oxidation is also a prerequisitef oxidative stress in obesity-related inflammatory liverisease.7,26 In fact, recent in vitro experiments show thathe accelerated oxidation of palmitate causes excess elec-ron flux in the mitochondrial respiratory chain, result-ng in increased free radical generation and hepatic insu-in resistance, and that the production of reactive oxygenpecies can be reduced by pharmacologic inhibition of FAxidation.9 In high-fat–fed animals, the up-regulation ofA oxidation and reactive oxygen species formation pre-edes the development of insulin resistance.27 In humaneings, �-hydroxybutyrate is more increased in progres-ive than benign obesity-related liver disease, and is ac-ompanied by mitochondrial structural defects.19 Unfor-unately, we did not address specific measurements ofxidative stress.

The hepatic flux of FA into the esterification pathwayppeared to be related to the rate of FA appearance in theirculation, consistent with previous findings in patientsith type 2 diabetes and high liver fat content.28 Because

ipolysis effectively was restrained by hyperinsulinemia inur obese subjects, as previously shown in subjects withimilar BMI and normal plasma FA levels,29 the hepaticA esterification and labeled TG release rates were pre-erved. Notably, the possibility of the liver to compensateor adipose tissue insulin resistance via the up-regulationf FA oxidation is limited because this process is alreadyery active in the healthy organ.30,31 Thus, the increase inepatic FA oxidation in our obese individuals suggests arogressive exhaustion of its buffering capability, to beollowed eventually by the diversion of FA into the ester-fication pathway. This mechanism may contribute, fos-ered by a worsening of FA release from fat, to theccumulation of lipids in the organ and to the patho-enesis of liver steatosis in more severe obesity. Mea-urements of liver fat content in 9 subjects showed aonsignificant tendency for higher hepatic uptake andsterification rates in a subgroup of 4 with liver fatontent greater than 6.5% (data not shown). Esterifica-ion may have been best evaluated during a meal chal-enge, but fasting studies were performed to best picturehe oxidative pathway and because of the steady-stateequirements of the current and traditional PET kinetic

odeling approach.In conclusion, the current methodology is able to non-

nvasively describe the fate of FA in the liver with satis-actory accuracy and detail. We show that a higher FAxidation is the first metabolic change in the liver ofbese individuals, occurring in the context of adiposeissue insulin resistance and increased FA flux from vis-eral adipose depots. This hepatic response is maladap-ive to the extent that a greater FA oxidation presumably

esults in the overproduction of reactive oxygen species

lready at this prodromal stage of obesity-related meta-olic syndrome.

Supplementary Material

Note: To access the supplementary materialccompanying this article, visit the online version ofastroenterology at www.gastrojournal.org, and at doi:0.1053/j.gastro.2010.05.039.

References

1. Charlton M. Nonalcoholic fatty liver disease: a review of currentunderstanding and future impact. Clin Gastroenterol Hepatol2004;2:1048–1058.

2. Petta S, Muratore C, Craxi A. Non-alcoholic fatty liver diseasepathogenesis: the present and the future. Dig Liver Dis 2009;41:615–625.

3. Brownlee M. Biochemistry and molecular cell biology of diabeticcomplications. Nature 2001;414:813–820.

4. Kotronen A, Seppala-Lindroos A, Bergholm R, et al. Tissue spec-ificity of insulin resistance in humans: fat in the liver rather thanmuscle is associated with features of the metabolic syndrome.Diabetologia 2008;51:130–138.

5. Yamaguchi K, Yang L, McCall S, et al. Inhibiting triglyceridesynthesis improves hepatic steatosis but exacerbates liver dam-age and fibrosis in obese mice with nonalcoholic steatohepatitis.Hepatology 2007;45:1366–1374.

6. Choi SS, Diehl AM. Hepatic triglyceride synthesis and nonalco-holic fatty liver disease. Curr Opin Lipidol 2008;19:295–300.

7. Fromenty B, Robin MA, Igoudjil A, et al. The ins and outs ofmitochondrial dysfunction in NASH. Diabetes Metab 2004;30:121–138.

8. Yang S, Zhu H, Li Y, et al. Mitochondrial adaptations to obesity-related oxidant stress. Arch Biochem Biophys 2000;378:259–268.

9. Nakamura S, Takamura T, Matsuzawa-Nagata N, et al. Palmitateinduces insulin resistance in H4IIEC3 hepatocytes through reac-tive oxygen species produced by mitochondria. J Biol Chem 2009;284:14809–14818.

0. Nielsen S, Guo Z, Johnson CM, et al. Splanchnic lipolysis inhuman obesity. J Clin Invest 2004;113:1582–1588.

1. Knuuti J, Takala TO, Nagren K, et al. Myocardial fatty acid oxida-tion in patients with impaired glucose tolerance. Diabetologia2001;44:184–187.

2. Arai T, Wakabayashi S, Channing MA, et al. Incorporation of[1-carbon-11]palmitate in monkey brain using PET. J Nucl Med1995;36:2261–2267.

3. Bergmann SR, Weinheimer CJ, Markham J, et al. Quantitation ofmyocardial fatty acid metabolism using PET. J Nucl Med 1996;37:1723–1730.

4. Klein S, Young VR, Blackburn GL, et al. Palmitate and glycerolkinetics during brief starvation in normal weight young adult andelderly subjects. J Clin Invest 1986;78:928–933.

5. Iozzo P, Jarvisalo MJ, Kiss J, et al. Quantification of liver glucosemetabolism by positron emission tomography: validation study inpigs. Gastroenterology 2007;132:531–542.

6. Rigazio S, Lehto HR, Tuunanen H, et al. The lowering of hepaticfatty acid uptake improves liver function and insulin sensitivitywithout affecting hepatic fat content in humans. Am J PhysiolEndocrinol Metab 2008;295:E413–E419.

7. Akaike H. [Data analysis by statistical models]. No To Hattatsu1992;24:127–133.

8. Natali A, Buzzigoli G, Taddei S, et al. Effects of insulin on hemo-dynamics and metabolism in human forearm. Diabetes 1990;39:

490–500.

1

2

2

2

2

2

2

2

2

2

2

3

3

R

o52

C

F

FhCPs((FsMw

CLIN

ICA

L–LIVER

,PA

NCREA

S,A

ND

BILIA

RY

TRA

CT

856 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

9. Sanyal AJ, Campbell-Sargent C, Mirshahi F, et al. Nonalcoholicsteatohepatitis: association of insulin resistance and mitochon-drial abnormalities. Gastroenterology 2001;120:1183–1192.

0. Lewis GF, Uffelman KD, Szeto LW, et al. Interaction between freefatty acids and insulin in the acute control of very low density lipoproteinproduction in humans. J Clin Invest 1995;95:158–166.

1. Malmstrom R, Packard CJ, Caslake M, et al. Effects of insulin andacipimox on VLDL1 and VLDL2 apolipoprotein B production innormal subjects. Diabetes 1998;47:779–787.

2. Durrington PN, Newton RS, Weinstein DB, et al. Effects of insulinand glucose on very low density lipoprotein triglyceride secretionby cultured rat hepatocytes. J Clin Invest 1982;70:63–73.

3. Iozzo P, Turpeinen AK, Takala T, et al. Defective liver disposal offree fatty acids in patients with impaired glucose tolerance. J ClinEndocrinol Metab 2004;89:3496–3502.

4. Brady LJ, Brady PS, Romsos DR, et al. Elevated hepatic mito-chondrial and peroxisomal oxidative capacities in fed and starvedadult obese (ob/ob) mice. Biochem J 1985;231:439–444.

5. Cook GA, Gamble MS. Regulation of carnitine palmitoyltrans-ferase by insulin results in decreased activity and decreasedapparent Ki values for malonyl-CoA. J Biol Chem 1987;262:2050–2055.

6. Day CP, James OF. Steatohepatitis: a tale of two “hits”? Gastro-enterology 1998;114:842–845.

7. Matsuzawa-Nagata N, Takamura T, Ando H, et al. Increasedoxidative stress precedes the onset of high-fat diet-induced insu-lin resistance and obesity. Metabolism 2008;57:1071–1077.

8. Kotronen A, Juurinen L, Tiikkainen M, et al. Increased liver fat,impaired insulin clearance, and hepatic and adipose tissue insu-lin resistance in type 2 diabetes. Gastroenterology 2008;135:122–130.

9. Bickerton AS, Roberts R, Fielding BA, et al. Adipose tissue fattyacid metabolism in insulin-resistant men. Diabetologia 2008;51:

1466–1474. T

0. Havel RJ, Kane JP, Balasse EO, et al. Splanchnic metabolism offree fatty acids and production of triglycerides of very low densitylipoproteins in normotriglyceridemic and hypertriglyceridemic hu-mans. J Clin Invest 1970;49:2017–2035.

1. Muller MJ. Hepatic energy and substrate metabolism: a possiblemetabolic basis for early nutritional support in cirrhotic patients.Nutrition 1998;14:30–38.

Received February 23, 2010. Accepted May 20, 2010.

eprint requestsAddress requests for reprints to: Patricia Iozzo, MD, PhD, Institute

f Clinical Physiology, National Research Council, Via Moruzzi 1,6124, Pisa, Italy. e-mail: patricia.iozzo@ifc.cnr.it; fax: 39 050 315166.

onflicts of interestThe authors disclose no conflicts.

undingThis work is part of the project “Hepatic and Adipose Tissue and

unctions in the Metabolic Syndrome” (HEPADIP,ttp://www.hepadip.org/), which is supported by the Europeanommission as an Integrated Project under the 6th Frameworkrogramme (contract LSHMCT-2005-018734). The study wasupported further by grants from the Finnish Diabetes FoundationP.I.), European Foundation for the Study of Diabetes (EFSD)/Eli-Lillyfellowship to P.I.), Sigrid Juselius Foundation (P.I.), Academy ofinland (P.N.), and the Finnish Cancer Institute (salary to A.R.). Thetudy was conducted within the “Finnish Centre of Excellence inolecular Imaging in Cardiovascular and Metabolic Research,”hich was supported by the Academy of Finland, University of

urku, Turku University Hospital, and Åbo Akademi University.

wwmoevmEF

fAscbbcgacmeppflttrt

sipmgiwtfe1

i1wpptats3m

lpeed

fsgptsemtPwtso

siofcmabbotnoB(

picSoetvassptst

September 2010 HEPATIC FATTY ACID METABOLISM 856.e1

Supplementary Materials and Methods

Validation studies were conducted in 6 pigs, inhich arterial, portal, and hepatic vein catheterizationas performed to compare PET derived with directlyeasured liver FA metabolism. The imaging methodol-

gy was transferred to 15 human studies, in which theffects of obesity were explored. The protocols were re-iewed and approved by the Ethical Committee for Ani-al Experiments of the University of Turku and the

thics Committee of the Hospital District of Southwest-inland for human studies.

Animal StudiesAnesthetized pigs (29 –31 kg) were studied during

asting (n � 3) or euglycemic hyperinsulinemia (n � 3).ccess to food was withdrawn at least 16 hours before the

tudy. Animals were anesthetized with ketamine and pan-uronium, and mechanically ventilated via tracheal intu-ation with normal room air (regulated ventilation, 16reaths/min). Catheters were placed in the jugular vein,arotid artery, and femoral vein, for the administration oflucose, insulin, and 11C-palmitate, and for sampling ofrterial blood. Splanchnic vessels were accessed by sub-ostal incision; after dissection of the hepatogastric liga-ent, purse string sutures were allocated to allow cath-

ter insertion via a small incision in the hepatic andortal vein. Doppler flow-probes were placed around theortal vein, assumed to represent 83% of liver bloodow.1 Surgical access was closed, and distal catheter ex-remities were secured to the abdominal surface to avoidip displacement. Vital signs, blood pressure, and heartate were monitored throughout the study. Animals wereransferred to the PET unit (Supplementary Figure 1).

PET imaging in animals. An ECAT 931-08/12canner (CTI, Inc, Knoxville, TN) was used. In the hyper-nsulinemic experiments, insulin was infused in arimed-continuous fashion at a rate of 5.0 mU � kg�1 �in-1, and euglycemia was maintained by a variable 10%

lucose infusion; in fasting studies, saline was infusednstead. A 20-minute transmission scan was performedith a removable Geranium-68 ring source for attenua-

ion correction. Synthesis of 11C-palmitate was per-ormed as previously described.2 After 60 minutes hadlapsed from the start of the insulin or saline infusion,1C-palmitate (371 � 67 MBq) was injected and dynamicmaging was performed for 50 minutes (20 frames, 8 �5, 2 � 30, 2 � 120, 1 � 180, 6 � 300, 1 � 600 s). Bloodas drawn simultaneously from the carotid artery andortal vein at each imaging time frame to measure totallasma radioactivity for computation of the input func-ion, and the carotid artery and hepatic and portal veinst 10 and 40 minutes for the assessment of tracer parti-ioning between the FA, TG, phospholipid, and water-oluble pool, as previously reported,3 and at 3, 5, 10, 15,0, and 50 minutes in the 11C-palmitate scan for the

easurement of 11C-CO2. Unlabeled FA, glucose, and TG f

evels were measured at baseline and during each scan, byreviously described procedures.4 At the end of eachxperiment, animals were killed and the whole liver wasxplanted. The volume of the organ was assessed by waterisplacement.

Image processing. All sinograms were correctedor tissue attenuation, dead time, and decay, and recon-tructed through standard algorithms. Large circular re-ions of interest were placed on 3 to 4 consecutive imagelanes in the right lobe of the liver for hepatic radioac-ivity measurements. The model input function, repre-enting the concentration of tracer available for liverxtraction from arterial and portal venous plasma, waseasured directly in blood samples in animals. The in-

act fraction of 11C-palmitate was used as input function.lasma radioactivity levels in the artery and portal veinere combined into one input function, resulting from

he relative contributions of the 2 vessels to liver perfu-ion, which were set at 17% and 83%, respectively, basedn our previous determinations.1

Human StudiesThe evaluation in human beings included 15

tudies, 9 of which were performed ad hoc to carefullymage the whole liver and visualize the portal vein, as partf the input function needed to calculate FA metabolismrom imaging data. These 9 studies were first used toonfirm relationships with circulating markers of liveretabolism, and to validate the extension of the current

pproach to studies lacking portal vein information, toe used in the remaining 6 human studies. The latter hadeen performed previously to evaluate myocardial metab-lism5; thus, images included only the upper portion ofhe liver. Human subjects were divided into obese andonobese based on a BMI cut-off level (30 kg/m2); thene patient with full-blown metabolic syndrome but aMI less than 30 kg/m2 was included in the obese group

n � 8).PET imaging in human beings. Each study was

erformed after an 8- to 10-hour overnight fast. PETmages were obtained by the GE PET-CT scanner (Dis-overy VCT; GE Medical systems, Milwaukee, WI), theiemens HR� PET scanner (Siemens, Inc, Knoxville, TN),r the ECAT 921/08-12 tomograph (CTI, Inc). Two cath-ters were inserted, one in an antecubital vein for injec-ion of 11C-palmitate, another in the opposite antecubitalein for blood sampling. Dynamic imaging of the upperbdominal area, lasting 50 minutes and including 20ampling frames, as described in the Animal Studiesection, was performed after administration of the radio-harmaceutical. During the scan, a heating pad was usedo arterialize venous blood at the sampling site and bloodamples were withdrawn at 10, 20, 30, 40, and 50 minuteso discount the image-derived input function for the

raction of metabolized radiopharmaceutical.3

lsiotttafurtp6ftc

srmr

bahocbtHtot(F2tc

hahwmagptwolm

atpw

nitdT(ispeset

ai

wctttw

iuIm((t

856.e2 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

Image processing. Image reconstruction and de-ineation of regions of interest was performed as de-cribed in the Animal Studies section. In 9 human stud-es, in which CT and image quality allowed identificationf the portal vein, small regions of interest were drawn inhis vessel and in the left ventricular cavity of the heart,o obtain radioactivity concentrations in portal and ar-erial blood, respectively, to be converted into plasmactivity through the hematocrit. Because the dual-inputunction is not always resolvable from human images, wesed a mathematic model to derive average kinetic pa-ameters that would permit its estimation from the ar-erial input function, as shown in the flowchart in Sup-lementary Figure 2. This input function was applied incontrol individuals. Subsequent data analysis was per-

ormed with either the dual (measured and estimated) orhe single arterial input function for quality and resultsomparisons.

Magnetic resonance studies. In the 9 humanubjects enrolled prospectively for this study, magneticesonance imaging and spectroscopy were performed to

easure intra-abdominal fat masses and liver fat content,espectively, as previously described.6

Sample AnalysesAssay of labeled 11C-CO2. In the animal studies,

lood samples (2 mL) were collected into tubes withoutny additives from the carotid artery, portal vein, andepatic vein during 11C-palmitate PET imaging. Aliquotsf blood (0.6 mL) were transferred immediately to tubesontaining 0.5 mol/L NaOH (2.4 mL) to fix CO2 asicarbonate, capped quickly, and briefly mixed; and toubes containing isopropanol (1.8 mL) and 0.5 mol/LCl (0.6 mL) to prevent clotting and enabling evapora-

ion of CO2, briefly mixed and then bubbled with argonr nitrogen for 10 minutes at 80°C. Radioactivity in bothubes was measured with an automatic gamma counter1480 Wizard 3” Gamma Counter; EG&G Wallac, Turku,inland) cross-calibrated with a dose calibrator (VDC-02; Veenstra Instruments, Joure, The Netherlands), andhe amount of radioactivity associated with 11C-CO2 wasalculated accordingly.

Assay of labeled lipid fractions. The procedureas been reported previously in detail, and metabolitenalysis and image quality assessments in these animalsave been reported previously.3 Briefly, blood samplesere collected in dry syringes, transferred to ethylenedia-inetetraacetic acid– coated tubes, mixed, and immedi-

tely stored in ice. After separation of plasma into or-anic and water components, as described,3 the aqueoushase was counted, and the organic phase was passedhrough a solid-phase extraction (SPE-NH2) column,hich had been activated previously by elution of 3 mLf heptane. Separation of triglycerides, FA, and phospho-

ipids was obtained by subsequent column elution with 3

L heptane:isopropanol (1:2 vol/vol), 4 mL 2% acetic

cid in diethyl ether, and 3 mL methyl alcohol, respec-ively. The same procedure was performed in blood sam-les from human studies. Radioactivity in each fractionas measured in a well counter.

Kinetic AnalysesCompartmental modeling of liver 11C-palmitate ki-

etics. Compartmental modeling describes each processnvolved in FA handling by the liver, including inwardransport, release of labeled 11C-products owing to oxi-ation of palmitate, incorporation of 11C-palmitate intoGs, and the subsequent partial release of labeled TGs

11C-TG) into the circulation. The model, which is shownn Figure 1A, consists of 3 tissue compartments, repre-enting free 11C-palmitate, 11C-palmitate bound in com-lex lipids, and 11C-oxidative breakdown products. Thexchange of radioactivity across compartments is de-cribed by 5 rate-constant terms, K1– k5. The differentialquations describing changing tracer concentrations inhe model compartments are as follows:

dC1/dt � Ca�t�xK1 � C2xk2 � C3xk3 (Eq. 1)

dC2/dt � C1xk2 � C2xk4 (Eq. 2)

dC3/dt � C1xk3 � C3xk5 (Eq. 3)

The total concentration of 11C-palmitate in the liver atgiven time, C(t), can be defined by the sum of the tracer

n each compartment and in the tissue blood fraction.

C�t� � �1 � Vb� x�C1�t� � C2�t� � C3�t�� � Cb�t�xVb

(Eq. 4)

here Ca is the metabolite-corrected plasma tracer con-entration, Cb(t) is the total blood radioactivity concen-ration, Vb represents the fraction of blood within theissue region of interest, and C1–3 is the concentration ofracer in each tissue compartment. The rate constant k5

as fixed to be equal to k3, as suggested.7

The system of differential equations was solved numer-cally, by using the commercial software SAAM II (Sim-lation Analysis and Modeling II, version 1.2.1; SAAM

nstitute, Seattle, WA). The rate constants obtained (1/in) were used together with plasma FA concentrations

CFA, �mol/mL) to derive substrate steady-state flux rates�mol/mL/min), according to Bergmann et al,8 by usinghe following formulas:

FA oxidation rate � CFAx�K1xk3�/�k2 � k3�(Eq. 5)

FA esterification rate � CFAx�K1xk2�/�k2 � k3�

(Eq. 6)

L

IF

1

2

3

4

5

6

7

8

September 2010 HEPATIC FATTY ACID METABOLISM 856.e3

abeled TG release rate � CFAx�K1xk2xk4�/

�3x�k3xk5 � k2xk4�� (Eq. 7)

FA uptake rate � esterification � oxidation rates

(Eq. 8)

n which labeled TG release takes into account the 3:1A:TG molar ratio.

References

. Iozzo P, Jarvisalo MJ, Kiss J, et al. Quantification of liver glucosemetabolism by positron emission tomography: validation study inpigs. Gastroenterology 2007;132:531–542.

. Padgett HC, Robinson GD, Barrio JR. [1-(11)C]palmitic acid: im-proved radiopharmaceutical preparation. Int J Appl Radiat Isot

1982;33:1471–1472.

. Guiducci L, Jarvisalo M, Kiss J, et al. [11C]palmitate kineticsacross the splanchnic bed in arterial, portal and hepatic venousplasma during fasting and euglycemic hyperinsulinemia. Nucl MedBiol 2006;33:521–528.

. Iozzo P, Turpeinen AK, Takala T, et al. Defective liver disposal offree fatty acids in patients with impaired glucose tolerance. J ClinEndocrinol Metab 2004;89:3496–3502.

. Knuuti J, Takala TO, Nagren K, et al. Myocardial fatty acid oxidationin patients with impaired glucose tolerance. Diabetologia 2001;44:184–187.

. Rigazio S, Lehto HR, Tuunanen H, et al. The lowering of hepaticfatty acid uptake improves liver function and insulin sensitivitywithout affecting hepatic fat content in humans. Am J PhysiolEndocrinol Metab 2008;295:E413–E419.

. de Jong HW, Rijzewijk LJ, Lubberink M, et al. Kinetic models foranalysing myocardial [(11)C]palmitate data. Eur J Nucl Med MolImaging 2009;36:966–978.

. Bergmann SR, Weinheimer CJ, Markham J, et al. Quantitation ofmyocardial fatty acid metabolism using PET. J Nucl Med 1996;37:

1723–1730.

Stft amic emission scan.

856.e4 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

upplementary Figure 1. Study design in the animal investigation. Times are specified on the right. The red thick line at top indicates frequenunctions. Labeled products were determined in samples collected simuhe PET transmission scan, followed by 11C-palmitate injection and dyn

he imaging or sampling procedures are listed on the left, and the samplingt blood sampling for the determination of the arterial and portal venous inputltaneously from the carotid artery and portal and hepatic veins. TR indicates

S

A

H

a

b

d

Sddimage-derived, dual-input function.

September 2010 HEPATIC FATTY ACID METABOLISM 856.e5

upplementary Table 1. Autocorrelation of Kinetic Parameters in the Animal and Human Studies

Absolute correlation coefficients (P values)

k4 K1 k3 k2

nimalsa

k4 1 0.51 (.30) 0.46 (.35) 0.71 (.11)K1 0.51 (.30) 1 0.19 (.72) 0.64 (.17)k3 0.46 (.35) 0.19 (.72) 1 0.75 (.08)k2 0.71 (.11) 0.64 (.17) 0.75 (.08) 1

uman beingsb

k4 1 0.21 (.45) 0.11 (.69) �0.01 (1.00)K1 0.21 (.45) 1 0.30 (.28) 0.04 (.90)k3 0.11 (.69) 0.30 (.28) 1 0.19 (.51)k2 �0.01 (1.00) 0.04 (.90) 0.19 (.51) 1

Data are from 6 animal studies; absolute coefficients are shown without sign owing to the absence of any correlation or tendency.Data are from 15 human subjects; absolute coefficients are shown without sign owing to the absence of any correlation or tendency. The result

upplementary Figure 2. Flowchart representation of the sequence of operations and compartmental model used to estimate parametersescribing the dual from the arterial input function in human beings (steps 1 and 2). These parameters subsequently were used to create newual-input functions and to estimate liver FA metabolism (step 4), the results of which were compared with those from the originally measured

id not change if obese and lean individuals were analyzed separately.

S

MVRV

NifwS

Spools and processes.

856.e6 IOZZO ET AL GASTROENTEROLOGY Vol. 139, No. 3

upplementary Table 2. Model-Derived Kinetic Parameters for Estimation of a Dual From an Arterial Input Function

Mean � SEM, kg (range)

Absolute correlation coefficients (P values)

k(0,2) k(2,1) k(2,3) k(3,2)

ean � SEM, min�1 2.26 � 0.12 1.96 � 0.12 0.40 � 0.08 0.86 � 0.17isceral fat 1.70 � 0.22 (.65–2.62) 0.28 (.47) 0.39 (.31) 0.33 (.39) 0.01 (.98)etrop fat 0.95 � 0.14 (.50–1.75) 0.21 (.58) 0.31 (.43) 0.51 (.17) 0.11 (.78)isceral � retrop fat 2.65 � 1.05 (1.17–4.37) 0.26 (.50) 0.36 (.34) 0.41 (.28) 0.04 (.93)

OTE. Mean values for fat masses (first column) and parameters (first raw) are given. Kinetic parameters are reported with the samedentification as Supplementary Figure 2. Data are from 9 human subjects in which fat masses were available, and the measured dual-inputunction (obtained from arterial left ventricular cavity and portal vein blood, identified via co-registration of CT and PET images) could be comparedith the estimated curves for validation purposes; absolute coefficients are shown without sign owing to the absence of any correlation.

upplementary Figure 3. Regression analyses between the model-predicted hepatic oxidative or nonoxidative metabolism of FA and the related

EM, standard error of the mean; retrop, retroperitoneal.