A Lipidomic Analysis of Nonalcoholic Fatty Liver Disease
Puneet Puri,1 Rebecca A. Baillie,3 Michelle M. Wiest,3 Faridoddin Mirshahi,1 Jayanta Choudhury,1 Onpan Cheung,1
Carol Sargeant,1 Melissa J. Contos,2 and Arun J. Sanyal1
The spectrum of nonalcoholic fatty liver disease (NAFLD) includes a nonalcoholic fatty liver
(NAFL) and nonalcoholic steatohepatitis (NASH). The specific types and amounts of lipids
that accumulate in NAFLD are not fully defined. The free fatty acid (FFA), diacylglycerol
(DAG), triacylglycerol (TAG), free cholesterol (FC), cholesterol ester, and phospholipid
contents in normal livers were quantified and compared to those of NAFL and NASH, and
the distribution of fatty acids within these classes was compared across these groups. Hepatic
lipids were quantified by capillary gas chromatography. The mean (nmol/g of tissue) DAG
(normal/NAFL/NASH: 1922 versus 4947 versus 3304) and TAG (13,609 versus 128,585
versus 104,036) increased significantly inNAFLD, but FFA remained unaltered (5533 versus
5929 versus 6115). There was a stepwise increase in the mean TAG/DAG ratio from normal
livers to NAFL toNASH (7 versus 26 versus 31,P< 0.001). There was also a similar stepwise
increment in hepatic FC (7539 versus 10,383 versus 12,863, P< 0.05 for NASH). The total
phosphatidylcholine (PC) decreased in both NAFL and NASH. The FC/PC ratio increased
progressively (0.34 versus 0.69 versus 0.71, P < 0.008 for both). Although the levels for
linoleic acid (18:2n-6) and �-linolenic acid (18:3n-3) remained unaltered, there was a
decrease in arachidonic acid (20:4n-6) in FFA, TAG, and PC (P < 0.05 for all) in NASH.
Eicosapentanoic acid (20:5n-3) and docosahexanoic acid (22:6n-3) were decreased in TAG
in NASH. The n-6:n-3 FFA ratio increased in NASH (P < 0.05). Conclusions: NAFLD is
associated with numerous changes in the lipid composition of the liver. The potential
implications are discussed. (HEPATOLOGY 2007;46:1081-1090.)
Nonalcoholic fatty liver disease (NAFLD) is acommon cause of chronic liver disease in NorthAmerica.1 NAFLD is associated with insulin re-
sistance and the metabolic syndrome.2,3 The clinical-his-
tologic spectrum of NAFLD extends from a nonalcoholic
fatty liver (NAFL) to nonalcoholic steatohepatitis
(NASH).4 Although NASH is distinguished from NAFL
by the presence of cytologic ballooning and inflamma-
tion, both conditions are characterized by a fatty liver.4,5
Thus, hepatic fat accumulation is the hallmark of
NAFLD.
The compositions of the lipids that accumulate in the
livers of subjects with NAFLD are not well characterized.
Most of the published literature has focused on triglycer-
ide accumulation as the key defect in NAFLD.6,7 How-
ever, it is not knownwhether there are substantial changes
in other lipid classes, such as cholesterol and specific phos-
pholipids (PLs). Although an increase in the n-6:n-3 fatty
acid ratio in total lipids in NAFLD has been described
recently,8,9 the distribution of these fatty acids within spe-
cific lipid classes has not been extensively characterized.
Given the important biological activities of many lipids,
such information could provide potential insights into the
pathophysiology of NAFLD and the metabolic syn-
drome.
A lipidomic approach was taken to quantify the major
lipid classes and the distribution of fatty acids within these
classes in the liver. The specific aims of the study were to
(1) quantify the absolute and relative amounts of free fatty
acids (FFAs), diacylglycerol (DAG), triacylglycerol
Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase;
BMI, body mass index; CE, cholesterol ester; CL, cardiolipin; DAG, diacylglycerol;
DGAT, diacylglycerol acyl transferase; EDTA, ethylene diamine tetraacetic acid; EFA,
essential fatty acid; FC, free cholesterol; FFA, free fatty acid; HDL, high-density lipid;
LCPUFA, long-chain polyunsaturated fatty acid; LDL, low-density lipid; LYPC, lyso-
phosphatidylcholine (lysolecithin); MUFA, monounsaturated fatty acid; NAFL, nonal-
coholic fatty liver; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic
steatohepatitis; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PL, phospho-
lipid; PS, phosphatidylserine; PUFA, polyunsaturated fatty acid; SFA, saturated fatty
acid; SM, sphingomyelin; TAG, triacylglycerol; TLC, thin-layer chromatography.
From the 1Division of Gastroenterology, Hepatology, and Nutrition, Department of
Internal Medicine, 2Department of Pathology, Virginia Commonwealth University
Medical Center, Richmond, VA; and 3Lipomics Technologies, West Sacramento, CA.
Received January, 2007; accepted April 6, 2007.
Supported by the following grants from the National Institutes of Health to A.J.S.:
K24 DK 02755-06, T32 DK 07150-31, and RO1 56331-05.
This is an original work and is not under consideration for publication elsewhere.
This work has been presented, in part, at the annual meeting of the European
Association for Study of the Liver in Vienna, Austria, 2006.
Address reprint requests to: Arun J. Sanyal, M.B.B.S., M.D., Professor of
Medicine, Pharmacology, and Pathology, Virginia Commonwealth University
Medical Center, MCV Box 980341, Richmond, VA 23298-0341. E-mail:
ajsanyal@hsc.vcu.edu; fax: (804) 828 2037.
Copyright © 2007 by the American Association for the Study of Liver Diseases.
Published online in Wiley InterScience (www.interscience.wiley.com).
DOI 10.1002/hep.21763
Potential conflict of interest: Dr. Wiest owns stock in Lipomics Technology.
1081
(TAG), free cholesterol (FC), cholesterol esters (CEs),
and PLs in normal livers and compare them to those of
NAFL and NASH and (2) compare the distribution of
fatty acids within each of these classes in these groups of
subjects. The goal was to define the changes in the hepatic
lipid composition in NAFLD and develop hypotheses
about the potential impact of such changes on the devel-
opment and progression of NASH.
Patients and Methods
Study Cohort
Consecutive subjects with elevated aminotransferases
were screened. All subjects had routine clinical, hemato-
logic, biochemical, and serologic evaluations. The sub-
jects with a history of excessive alcohol use (�20 g/day for
males and�10 g/day for females) or other causes of liver
disease (viral hepatitis B, viral hepatitis C, primary biliary
cirrhosis, sclerosing cholangitis, autoimmune hepatitis,
hemochromatosis, Wilson’s disease, �1-antitrypsin defi-
ciency, and drug-induced liver disease) were excluded.
NAFLD was suspected by the presence of abnormal liver
enzymes without evidence of other liver diseases or radio-
logic evidence of a fatty liver.
Subjects with suspected NAFLD who were undergo-
ing a liver biopsy to further evaluate their liver disease
were considered for this study. Also, subjects without
symptoms or signs of liver disease and normal alanine
aminotransferase and sonogram undergoing abdominal
surgery for unrelated reasons served as a normal control
group. After an overnight fast, a core biopsy was obtained
in all cases with either a 15-gauge Microvasive gun or a
16-gauge Klatskin needle. The presence of NAFLD was
diagnosed according to standard clinical criteria.10,11 On
the basis of the liver histology, 3 groups were studied: (1)
normal histology controls, (2) NAFL, and (3) NASH.
Subjects with bridging fibrosis or cirrhosis were excluded.
The study was performed according to the Virginia Com-
monwealth University regulations for the protection of
human research subjects after the protocol was reviewed
and approved by the institutional review board.
Reagents
All chromatography solvents were obtained from
Fisher Scientific (Pittsburgh, PA). Silica Gel 60 thin-layer
chromatography (TLC) plates (10 �20 cm) were ob-
tained from E. Merck (Darmstadt, Germany). Ethylene
diamine tetraacetic acid (EDTA) and butylated hydroxy-
toluene were obtained from Sigma Chemical Co. (St.
Louis, MO). Fatty acid methyl ester and internal PL stan-
dards were obtained from Nu-Chek-Prep (Elysian, MN)
and Avanti Polar Lipids (Alabaster, AL), respectively.
Lipid Profiling
The hepatic lipid profiles were analyzed as previously
described12 and noted in the following.
Extraction and TLC. Lipids were extracted from
5-10 mg of liver tissue in the presence of internal stan-
dards by Folch’s method with chloroform/methanol (2:1
vol/vol).13 Individual lipid classes from each extract were
separated by preparative chromatography, as described
previously.14 Briefly, TLC plates were impregnated with 1
mM EDTA (pH 5.5) and washed by ascending develop-
ment.15 Sample extracts were dried under nitrogen and
spotted onto EDTA-impregnated TLC plates. Two TLC
standard lanes consisting of authentic phosphatidylcho-
line (PC), phosphatidylethanolamine (PE), cardiolipin
(CL), and FFA were spotted on the outside lanes of the
TLC plate as reference samples. The chloroform/metha-
nol/acetic acid/water (100:67:7:4 by volume) mobile
phase employed for the separation of PL classes was a
modification of the solvent system described previously.16
For the separation of lipid classes [total PL, FFA, TAG,
DAG, FC, and CE], a petroleum ether/diethyl ether/ace-
tic acid (80:20:1 by volume) solvent system was em-
ployed.17 These methods were initially validated in tissue
samples of various sizes against internal standards, and
accurate data could be obtained from as little as 3-4 mg of
liver.
Isolation and Methylation of the Lipid Classes.
Lipid classes and individual PL classes were identified by
comparison with the authentic standards chromato-
graphed in the reference lanes. Lipid fractions were trans-
esterified in 3 N methanolic HCl under an N2
atmosphere at 100°C for 45 minutes in a sealed vial. The
resulting fatty acid methyl esters were extracted with hex-
ane containing 0.05% butylated hydroxytoluene and pre-
pared for gas chromatography through the sealing of the
hexane extracts under nitrogen.
Fatty Acid Analysis. Fatty acid methyl esters were
separated and quantified by capillary gas chromatography
with a Hewlett-Packard (Wilmington, DE) gas chro-
matograph (model 6890) equipped with a 60-m DB-23
capillary column (J&W Scientific, Folsom, CA), a flame-
ionization detector, and Hewlett-Packard ChemStation
software.
Data Analysis
Summary data for the fatty acid classes and mole per-
cent (percentage of each fatty acid of the total fatty acids
within each lipid class) were calculated. The results were
expressed as means� SEM. A Kruskal-Wallis analysis of
variance with a post hoc multiple comparison procedure
was used for across-group comparisons. A Student t test
was used only when 2 specific groups of normally distrib-
1082 PURI ET AL. HEPATOLOGY, October 2007
uted data were being compared. A P value of 0.05 or less
was considered significant.
Results
Clinical, Biochemical, and Histologic Profile
A total of 9 subjects in each group were studied (Table
1). The 3 groups were similar with respect to gender,
ethnicity, age, body mass index, fasting blood sugar, gly-
cosylated hemoglobin, and hepatic synthetic functions.
Subjects with NASH had higher aspartate aminotransfer-
ase and alanine aminotransferase levels that approached
but did not reach significance. Although the total choles-
terol was higher in subjects with NAFL and NASH than
controls (P � 0.05 for both), there were no significant
differences in high-density lipid cholesterol or low-den-
sity lipid cholesterol. Subjects with NAFL had isolated
steatosis (mean grade� SEM: 2.1� 0.6) or steatosis with
mild inflammation (meanNAFLD activity score� SEM:
3.2� 0.3), whereas those with NASH had a mean steato-
sis score � SEM of 1.8 � 0.4 and an NAFLD activity
score� SEM of 5.1� 0.4 (P � 0.03 versus NAFL).
Total Hepatic Lipid Content (Table 2)
The total hepatic lipid content was markedly increased
in NAFL and NASH (P� 0.001); this was driven mainly
by increased TAG content (P� 0.001). Similarly, DAG
was also increased significantly (P� 0.03 for both NAFL
and NASH). Compared to that of normal controls, the
TAG/DAG (product/precursor) ratio was significantly
increased in both NAFL and NASH (7 versus 26 versus
31, P� 0.001 for both). The n-6 fatty acid content in the
total lipids increased from the controls to NAFL and
NASH (mean � SEM: 4131 � 210 versus 6424 � 605
versus 8449� 1012 nmol/gm, P� 0.03 forNASH versus
the controls) and was associated with a significantly
higher total n-6:n-3 ratio in NAFL and NASH (P� 0.05
for both versus the controls). However, the FFA content
did not increase in eitherNAFL orNASH. Also, there was
a stepwise increment in FC from normal to NAFL to
NASH (P � 0.05 for the control versus NASH). This
was, however, not accompanied by an increase in CEs in
either NAFL or NASH.
The total hepatic PL content was not significantly dif-
ferent across the 3 groups (mean � SEM: controls �
49,889 � 3220 nmol/g of tissue, NAFL � 42,671 �
3714 nmol/g of tissue, and NASH � 49,614 � 2599
nmol/g of tissue, P� not significant). However, despite a
significant increase in the total lipids and FC, the total PC
content was decreased (P � 0.03 for NAFL versus the
control). This was accompanied by an increase in lyso-
phosphatidylcholine (LYPC; i.e., lysolecithin) in subjects
Table 1. Baseline Characteristics of the Study Population
Parameter
Control
(n � 9)
NAFL
(n � 9)
NASH
(n � 9)
Gender F/M 7/2 6/3 6/3
Caucasian/African American 7/2 8/1 9/0
Mean age (years) 46.6� 3.8 48.5� 4.8 47� 3.2
BMI (kg/m2) 34.5� 4.3 37.5� 5.2 34� 1.8
Fasting glucose (mg/dl) 87.5� 4.9 119.2� 39.6 123.3� 29.3
Glycosylated hemoglobin
(%) 5.2� 0.4 5.8� 0.1 5.8� 0.1
AST (normal range:
0-65 U/l) 42� 16.6 36.8� 5 77.4� 14.2
ALT (normal range:
0-65 U/l) 55.7� 27.4 49� 10.2 119.4� 25.8
Alkaline phosphatase (U/l) 82.4� 9.1 113.4� 22.8 111� 13.9
Total bilirubin (mg/dl) 0.6� 0.1 0.5� 0.1 0.5� 0.1
Albumin (g/dl) 4.1� 0.1 4� 0.3 4� 0.1
Platelets (1000/mm3) 255.5� 27.9 281.3� 40.5 233� 23.3
Total cholesterol (mg/dl) 166.5� 10.5 230.6� 13.8* 234� 11.2*
HDL cholesterol (mg/dl) 57.5� 8.5 40.2� 6.3 50.5� 2.1
LDL cholesterol (mg/dl) 87� 4 127� 22.6 99� 3.8
Triglycerides (mg/dl) 220� 22 397.8� 135.2 274� 19.4
Steatosis grade 0 2.1� 0.6 1.8� 0.4
NAFLD activity score 0 3.2� 0.3 5.1� 0.4†
Fibrosis stage 0 1.3� 0.3 1.8� 0.4
The data are expressed as the mean � SEM. ALT indicates alanine amino-
transferase; AST, aspartate aminotransferase; BMI, body mass index; HDL, high-
density lipid; LDL, low-density lipid; NAFL, nonalcoholic fatty liver; NAFLD,
nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis.
*P � 0.05 versus the control (analysis of variance).
†P � 0.05 versus NAFL (t test).
Table 2. Hepatic Lipid Content of the Study Groups: Total
and Individual Lipid Classes
Parameter
Control
(n � 9)
NAFL
(n � 9)
NASH
(n � 9)
Total lipids 15,978� 1,157 49,991� 10,718* 43,658� 6,162*
Total SFA 7,189� 586 22,889� 5,328* 18,756� 3,003*
Total MUFA 3,527� 258 19,616� 4,937 14,819� 2,204*
Total PUFA 5,104� 347 7,328� 606* 9,822.62� 1,094*
Total n-3 949� 153 862� 49 1,336� 180
Total n-6 4,131� 210 6,424� 605 8,449� 1,012*
n-6/n-3 ratio 4.8� 0.4 7.6� 0.8* 6.9� 0.9*
FFA 5,533� 1,210 5,929� 1,487 6,115� 584
DAG 1,922� 430 4,946� 1,232* 3,304� 473*
TAG 13,609� 2,300 128,585� 34,201* 104,035� 20,451*
TAG/DAG ratio 7 26* 31*
CE 7,589� 1,548 8,522� 1,846 7,774� 808
FC 7,538� 718 10,382� 1,079 12,862� 1,707*
CL 4,447� 620 5,023� 585 6,063� 902
LYPC 1,936� 308 1,800� 505 2,239� 475*
PC 20,321� 1,098 15,322� 1,247* 16,874� 1,319
PE 14,599� 813 11,456� 494* 14,828� 1,359
PS 4,114� 577 5,295� 1,774 4,199� 706
SM 4,194� 587 3,773� 590 5,684� 685
All values are expressed as the mean (nmol/g of tissue) � SEM. CE indicates
cholesterol ester; CL, cardiolipin; DAG, diacylglycerol; FC, free cholesterol; FFA,
free fatty acid; LYPC, lysophosphatidylcholine; MUFA, monounsaturated fatty acid;
NAFL, nonalcoholic fatty liver; NASH, nonalcoholic steatohepatitis; PC, phosphati-
dylcholine; PE, phosphatidylethanolamine; PUFA, polyunsaturated fatty acid; PS,
phosphatidylserine; SFA, saturated fatty acid; SM, sphingomyelin; TAG, triacyl-
glycerol.
*P � 0.05 versus the control (analysis of variance).
HEPATOLOGY, Vol. 46, No. 4, 2007 PURI ET AL. 1083
with NASH (P � 0.05). There was also a trend for de-
creased PE that was significant only forNAFL (P� 0.05);
in contrast, phosphatidylserine (PS) levels remained un-
changed. Although the hepatic content of sphingomyelin
(SM) and CLs increased, they did not meet levels of sig-
nificance.
Fatty Acid Composition of Hepatic FFAs (FFA;
Table 3)
There was a trend for a progressive decrease from con-
trols to NAFL and then NASH for n-6 and n-3 polyun-
saturated fatty acids (PUFAs) within hepatic FFA. The
content of linoleic acid (18:2n-6), the starting point for
processing n-6 essential fatty acids (EFAs),18 was unal-
tered in both NAFL and NASH. However, there was a
trend toward progressive depletion of �-linolenic acid
(18:3n-6), which is immediately downstream of linoleic
acid,18,19 from the control to NAFL to NASH (P� 0.01
versus NASH). This was also seen with arachidonic acid
(20:4-n6), which is further downstream of �-linolenic
acid, with a significant decrease in NASH (P � 0.05
versus the controls). The product/precursor ratio for the
n-6 pathway (20:4n-6:18:2n-6) trended downward,
reaching significance for NASH (P � 0.03 versus the
controls). The levels of mead acid (20:3n-9), which are
typically increased in EFA deficiency,20 remained unal-
tered in NAFL and NASH.
The �-linolenic acid (18:3n-3), the starting point for
processing n-3 EFA,18 was unaltered. There was a trend
for a progressive decrease from the controls to NAFL to
NASH for eicosapentanoic acid (20:5n-3) and docosa-
hexanoic acid (22:6n-3), the downstream products of
�-linolenic acid (18:3n-3), which approached but did not
reach significance in NASH. The product/precursor ratio
for the n-3 pathway (20:5n-3:18:3n-3) also trended
downward, approaching significance for NASH.
Fatty Acid Composition of Hepatic DAG and TAG
Both DAG and TAG were significantly increased in
subjects with NAFL and NASH (Fig. 1A). There was a
trend for increased saturated fatty acids (SFAs) in DAG
and TAG; this was driven by increased palmitate (16:0)
but offset by decreased stearic acid (18:0). There was also
a trend for increased monounsaturated fatty acids (MU-
FAs; Fig. 1B), specifically oleic acid (18:1-n9), in DAG
and TAG for both NAFL and NASH.
In contrast, there was a significant decrease in PUFA
associated with DAG and TAG in NAFL and NASH
(Fig. 1B). The molar percentages of n-3 and n-6 fatty
acids in TAG were decreased in both NAFL and NASH
(Fig. 1C); however, the decrease in n-6 was less than that
in the n-3 fatty acids, resulting in a net significant increase
in the n-6:n-3 fatty acid proportions in TAG (Fig. 1D).
There was a significant depletion of arachidonic acid (20:
4n-6), a key n-6 fatty acid (Fig. 1E). Also, eicosapentanoic
acid (20:5n-3) and docosahexanoic acid (22:6n-3), the 2
downstream products of �-linolenic acid (18:3n-3) in the
n-3 pathway, were significantly depleted (Fig. 1F). These
changes were qualitatively similar to the changes seen in
the FFA and DAG pools.
Hepatic CE Fatty Acid Composition
The hepatic FC content increased progressively from
controls with normal histology to NAFL to NASH (P�
0.05 for NASH versus the control; Fig. 2A). The total CE
content was not, however, significantly changed in either
NAFL or NASH. There was a de-enrichment of SFA and
a mild, insignificant increase in MUFA, specifically oleic
acid. There was a significant enrichment of the CEs with
PUFA (Fig. 2B,C). Both n-6 and n-3 PUFAs increased in
NAFL and NASH, but the findings were significant only
for n-6 fatty acids (Fig. 2C). The overall n-6:n-3 ratio did
not change significantly (Fig. 2D). Although linoleic acid
(18:2n-6) was particularly enriched within the CEs in
both NAFL and NASH, the arachidonic acid levels were
not altered significantly (Fig. 2E). Although there was a
Table 3. Fatty Acid Composition of the FFA