De novo lipogenesis in humans: metabolic and regulatory aspects
Department of Nutritional Sciences, University of California at Berkeley, Berkeley, CA 94270-3104, USA
The enzymatic pathway for converting dietary carbohydrate (CHO) into fat, or de novo lipogenesis (DNL), is
present in humans, whereas the capacity to convert fats into CHO does not exist. Here, the quantitative
importance of DNL in humans is reviewed, focusing on the response to increased intake of dietary CHO.
Eucaloric replacement of dietary fat by CHO does not induce hepatic DNL to any substantial degree. Similarly,
addition of CHO to a mixed diet does not increase hepatic DNL to quantitatively important levels, as long as
CHO energy intake remains less than total energy expenditure (TEE). Instead, dietary CHO replaces fat in the
whole-body fuel mixture, even in the post-absorptive state. Body fat is thereby accrued, but the pathway of DNL
is not traversed; instead, a coordinated set of metabolic adaptations, including resistance of hepatic glucose
production to suppression by insulin, occurs that allows CHO oxidation to increase and match CHO intake. Only
when CHO energy intake exceeds TEE does DNL in liver or adipose tissue contribute signi®cantly to the whole-
body energy economy.
It is concluded that DNL is not the pathway of ®rst resort for added dietary CHO, in humans. Under most dietary
conditions, the two major macronutrient energy sources (CHO and fat) are therefore not interconvertible
currencies; CHO and fat have independent, though interacting, economies and independent regulation. The
metabolic mechanisms and physiologic implications of the functional block between CHO and fat in humans are
discussed, but require further investigation.
In this review, I will address the fate of surplus dietary
carbohydrate (CHO) in humans. More speci®cally, the
focus will be on conversion of CHO to fat, or de novo
lipogenesis (DNL), with the question framed in quantitative
terms: to what extent is surplus dietary CHO energy
converted to fat? The various ways in which CHO content
of the diet can be increased will be considered: increased
CHO that replaces dietary fat (high-CHO low-fat, eucaloric
diets); CHO added to a mixed diet, where CHO energy is
less than total energy expenditure (TEE) but total energy
intake exceeds TEE; and CHO consumption in excess of
TEE. This review will therefore focus on the upper limits
and consequences of increased CHO intake rather than on
the lower limits and consequences of insuf®cient fat intake.
Background and historical review
The enzymatic pathway of DNL is present in all organisms.
Knowledge concerning the genes and enzymes of DNL and
their regulation has advanced considerably (reviewed in
Bloch, 1977; Girard et al, 1994; Sul et al, 1993). Never-
theless, quantitative and regulatory aspects of DNL in
metabolic physiology remain controversial. It has been
widely presumed that DNL functions primarily as a sink
for storage of excess CHO energy and to a lesser extent for
the synthesis of structural, non-essential lipids (Table 1).
Indeed, many animals are well known to convert CHO to
fat (Lawes & Gilbert, 1886; Florkin & Stotz, 1977): pigs
fatten on a grain diet, for example, and bees convert honey
to wax. Although conversion of CHO to fat prior to
oxidation is believed to be thermogenically costly (using
ca. 28% of the energy content of CHO, Flatt, 1978;
Hellerstein et al, 1996), the pathway is known to have
regulated steps and is therefore presumed to play a role in
normal physiology. Despite considerable information about
the regulation of acetyl-CoA carboxylase, fatty acyl synthe-
tase, malonyl-CoA and other components of the enzymatic
pathway (Bloch, 1977; Girard et al, 1994), the quantitative
importance of DNL has remained an area of uncertainty
Most of this uncertainty can be attributed to limitations in
the methods available to address this question. Both indirect
and direct techniques for measuring DNL have been used.
Indirect approaches for assessing the role of DNL in
A number of indirect approaches have been applied to this
Comparison of fatty acid (FA) composition in adipose
tissue and diet
Over 30 y ago, Hirsch (1965) observed that adipose FA
composition in human subjects closely resembled that of
the Western diet. Individuals were also placed on con-
*Correspondence: MK Hellerstein, Department of Nutritional Sciences,
University of California at Berkeley, Berkeley, CA 94270-3104, USA.
Table 1 Overview of de novo lipogenesis (DNL) as a pathway
(A) Presumed functions of DNL in the organism:
Synthesis of structural lipids
Storage of surplus CHO energy as fat
(B) High thermogenic costs
(C) Enzymology and regulation of DNL:
Key regulatory node Acetyl-CoA carboxylase (ACC)
Malonyl-CoA as product of ACC and regulator of FA oxidation
(D) Quantitative importance of DNL uncertain
European Journal of Clinical Nutrition (1999) 53, Suppl 1, S53±S65
ß 1999 Stockton Press. All rights reserved 0954±3007/99 $12.00
trolled diets of different FA compositions, for 6 ± 12
months, to allow turnover of adipose FA depots. The
adipose FA composition changed to resemble that of the
new diet. These investigators concluded that the body adds
little endogenous FA to the dietary mix, that is, we are what
we eat with regard to whole body FA stores. The obvious
inference from these ®ndings is that DNL is not a quantita-
tively important pathway in humans, at least under condi-
tions of Western (high fat) diets.
Other explanations of these results can also be put
forward, however. It is possible that selectivity exists for
storage (esteri®cation) relative to oxidation (lipolysis and
entry into mitochondria) for FA in humans. If this were the
case, de novo FA could be synthesized and oxidized with-
out accumulating in tissues. Nor do static changes in FA
composition provide information about ¯uxes through path-
ways; for example, DNL=fat oxidation cycling could occur
at a very high or a very low rate and remain consistent with
these observations. Nevertheless, a similar principle (dilu-
tion of the essential FA linoleate in circulating lipids) has
been used recently as a method for estimating rates of DNL
(Petrek et al, 1997; Hudgins et al, 1996; see below).
Enzyme activities in tissues
The enzyme activities of the DNL pathway in adipose
tissue and liver have been used to estimate an upper
bound on in vivo lipogenesis. Most studies (Shrago et al,
1971; Weiss et al, 1986; Sjo
strom, 1973; Askanazi et al,
1980) have reported quite low maximal lipogenic rates in
human lipogenic tissues compared to rodents, for example
(Table 2). These studies might underestimate capacity for
inducing DNL, however, since most subjects were studied
under typical high-fat eucaloric conditions. Sjo
reported only modest stimulation of DNL enzymes in
humans after a high CHO=low fat diet (Sjo
During parenteral overfeeding some evidence has been
presented (Askanazi et al, 1980) for induction of DNL
enzymes in human adipose tissue, but the quantitative
importance has not been established.
Indirect calorimetry to measure net DNL in the whole body
The most widely used technique for assessing DNL in vivo
has been indirect calorimetry. The principles of gas
exchange to measure fuel selection have been reviewed
quier et al, 1987; Elia & Livesey, 1988). This
approach can be used to estimate net conversion of CHO to
fat, since DNL is the only metabolic process with an
RQ > 1.0. Although different estimates of the true RQ for
palmitate synthesis have been provided (Hellerstein et al,
quier et al, 1987), it is clear that DNL within a
biologic system will generate RQ > 1.0 and that RQ > 1.0
represents DNL. The major interpretative problem with
indirect calorimetry in this regard is that it measures net
DNL, not unidirectional ¯ux through the pathway. An NP
RQ > 1.0 indicates only that synthesis is greater than
oxidation of fat in the whole system during the period
that is sampled. Lipogenesis from CHO in one tissue could
be balanced by fat oxidation in another, for example, and
the resultant RQ (1.0) would not be distinguishable from
direct CHO oxidation (Tappy et al, 1995).
Nevertheless, indirect calorimetry has provided useful
information about the response to large CHO loads. A
number of studies (Acheson et al, 1982, 1984; Hellerstein
et al, 1991) have con®rmed that a single meal containing
large amounts of CHO energy (up to 500 g CHO) in
previously weight stable subjects does not cause NP RQ
to rise above 1.0. This has been interpreted as evidence
against a quantitatively important role for DNL in the day-
to-day storage of surplus CHO energy; also, storage as
glycogen was concluded to represent the fate of excess
dietary CHO. Several days of surplus energy intake as CHO
(Schwarz et al, 1995; Passmore and Swindells, 1963) also
does not induce much net DNL (RQ reaches 0.98 ± 1.01). In
contrast, studies of massive CHO overfeeding
(6000 kcal=d; 1500 g CHO) performed by Acheson et al
(1988) showed that after about three days, when a positive
whole CHO balance of 800 g had occurred, RQ rose well
above 1.0; these investigators calculated that maximal
whole-body glycogen storage capacity was ca. 700 ±
1000 g and needed to be exceeded before net DNL
became signi®cant, after which time substantial net DNL
occurred (for example, 150 g=d net fat synthesis after seven
days of overfeeding). These dietary studies were highly
unphysiologic, however, and do not clarify the physiologic
role of DNL under less extreme dietary stresses.
Stoichiometry (macronutrient intake vs changes in body
The least ambiguous indirect evidence for DNL would
come from stoichiometry: if the rate of accumulation of
adipose fat is greater than the intake of dietary fat, there
must have been at least that much DNL. The only study in
humans using this approach was the fascinating report of
the `Guru Walla' overfeeding (Pasquet et al, 1992), a
traditional fattening ritual in Cameroon wherein adolescent
boys are deliberately overfed to an extraordinary degree.
These young men ingest a high CHO diet (70%) containing
7000 kcal=d; they gain 12 kg fat over 10 weeks while eating
Table 2 Tissue enzyme activities in humans and rat
Human Rat Human Rat
Enzyme pr sc om pr sc om ep
PDH 1.5 16.4 Ð 2.0 Ð Ð Ð Ð 23.6
ATP citrate lyase 1.0 0.8 0.9 n.d. 1.5 73.1 33.5 36.8 Ð
FA synthase 0.8 6.4 1.0 0.9 1.0 52.2 Ð Ð 54.3
Measurements shown in milliunits per milligram of protein. Abbreviations: pr, perirenal; sc, subcutaneous; om, omental; ep, epididymal; n.d., not
detectable. Adapted from Shrago et al, 1971; Weiss et al, 1986. Data on human tissues were from subjects on typical Western diets (fat content up to 40%).
No differences were observed after a high-carbohydrate diet for three days in these studies.
De novo lipogenesis in humans
a total of only 4 kg fat. Clearly, under these exceptional
nutritional conditions where CHO intake (ca. 5000 kcal=d)
greatly exceeds TEE, surplus CHO must be converted to
fat, even though DNL was not directly measured. It must be
concluded then that DNL in the whole body at a quantita-
tively signi®cant rate is possible in humans. Many ques-
tions were not answered, however (Table 3): in what
tissue(s) did the DNL occur? Was there even greater
unidirectional DNL than apparent by net accumulation of
fat (was there a circadian rhythm of fat synthesis and
subsequent oxidation)? What was the threshold for total
CHO surplus after which DNL was induced? Are there
genetic differences for DNL between this ethnic group and
other human populations? And, do people ever voluntarily
ingest enough surplus CHO to induce signi®cant DNL; if
not, why not? The answer to this last question may provide
important insights into the overall regulation of energy
balance in humans, including the integration of food intake
with nutrient stores.
Direct measurement of DNL in vivo in humans using
The direct approach for measuring ¯ux through the DNL
pathway or any metabolic pathway in vivo generally
requires the use of isotopes. It is therefore necessary to
discuss brie¯y a few technical points concerning isotopic
measurement of DNL, before we can interpret physiologic
results using this approach. Measurements of biosynthetic
processes in general, not only DNL, but also protein
synthesis, gluconeogenesis, polysaccharide synthesis, poly-
nucleotide synthesis, etc., have presented substantial tech-
nical problems (reviewed in Hellerstein & Neese, 1992;
Waterlow et al, 1978; Katz, 1985; Dietschy & Brown,
1974). The central and most intractable of these problems
has been how to gain access to the degree of isotope
labeling of the true biosynthetic precursor pool, in order
to interpret isotope incorporation into the biosynthetic
product as a chemical ¯ux rate. For the case of lipogenesis,
Dietschy & Brown (1974) pointed out over 20 y ago that
errors in the measurement of cholesterol and FA synthesis
rates are unavoidable when using labeled acetate incorpora-
tion, because of variable, unpredictable and often extensive
intracellular dilution in the cytosolic acetyl-CoA pool.
Fortunately, over the past half-dozen years or so, two
methods have been developed that resolve this problem and
therefore allow reliable measurement of DNL. The two
methods are mass isotopomer distribution analysis (MIDA)
The MIDA technique has been described in detail else-
where (Hellerstein et al, 1991, 1996; Hellerstein & Neese,
1992; Neese et al, 1995; Faix et al, 1993; Hellerstein,
1995). In brief, this is a stable isotope mass spectrometric
method for measuring polymerization biosynthesis based
on the principles of combinatorial probabilities. The basic
concept is that polymers are composed of two or more
repeating monomeric subunits, which are assembled from
an intracellular biosynthetic precursor pool; and the quan-
titative pattern or distribution of labeled subunits in the
polymer population re¯ects the isotopic enrichment of the
monomeric pool, according to the binomial or multinomial
expansion. It follows that analysis of the labeling pattern
(the mass isotopomer distribution) in the polymer by
quantitative mass spectrometry will reveal the intracellular
precursor pool enrichment. Fractional biosynthesis can then
be calculated by application of the precursor-product rela-
tionship (Hellerstein & Neese, 1992). A numerical example
of combinatorial probabilities (Figure 1A) and a general
scheme of mass isotopomer distributions (Figure 1B) are
O (labeled water) technique (Table 4) attempts
to overcome the problem of metabolite compartmentaliza-
tion and differential dilution by using H
O, which is freely
permeable across almost all biological membranes (other
than distal renal tubule) and is well mixed in the body
(Leitch & Jones, 1993; Jones et al, 1995). Labeled hydro-
gen from H
O enters into biosynthetic pathways either by
exchange with solvent H
O or by speci®c incorporation of
H from NAD(P)H (Jungas, 1968). If the number of hydro-
gen atoms incorporated from H
O per mole of product is
known, the isotopic enrichment of
O from any
body ¯uid can be used to calculate fractional biosynthesis
by application of the precursor-product relationship. The
major complication here is determining the actual H:C ratio
present in a biosynthetic product in a particular tissue in a
given experiment (Selmer & Grunnet, 1976; Lee et al,
1994). The H:C ratio has been determined accurately by
use of MIDA (Lee et al, 1994), but this approach has not
been applied in humans. The existence of an isotope effect,
O particularly but for
O as well, represents
another potential experimental artifact. Nevertheless,
using literature values for H:C ratios, DNL has been
measured in humans by this method (Leitch & Jones,
1993; Jones et al, 1995).
It is important to mention what might seem an obvious
point: an investigator can only measure DNL in pools that
can be experimentally sampled. If DNL occurs in a
`hidden' location, that is, inaccessible to collection by the
investigator, it will not be measured. This caveat becomes
important for interpretation if there is a possibility of newly
synthesized TG being stored in the hepatic cytosolic pool
rather than secreted as VLDL-TG, for example, or when
adipose DNL might be occurring, since the enormous
adipose TG pool size might dilute out recent incorporation
of label and preclude measurement of DNL over the short-
term (see below).
Regulation of DNL in humans
Results from both MIDA (Hellerstein et al, 1996; Hudgins
et al, 1996; Hellerstein et al, 1991; Neese et al, 1995; Faix
et al, 1993; Hellerstein, 1995) and
(Leitch & Jones, 1993; Jones et al, 1995) agree that hepatic
DNL is a quantitatively minor pathway in men under
conditions of a normal Western (high fat) diet. Calculated
synthesis rate of palmitate or total non-essential FA based
on ¯ux through the circulating VLDL-TG pool is < 1±
2g=d, a small amount compared to the 50 ± 150 g fat in the
diet each day. These results con®rm prior conclusions from
indirect calorimetry (Tappy et al, 1995; Acheson et al,
Table 3 `Guru Walla' Model Ð Summary
(1) Implications: Substantial DNL can be induced in humans, under
extreme dietary conditions (CHO intake >> TEE)
(2) Unanswered questions:
Threshold for inducing signi®cant DNL?
Tissue site of DNL?
Diurnal cyclicity of DNL?
Is absolute DNL > net fat accumulation?
De novo lipogenesis in humans
1982, 1984; Hellerstein et al, 1991), comparisons of tissue
to diet FA composition, and in vitro enzyme measurements
(Shrago et al, 1971; Weiss et al, 1986; Sjostrom, 1973)
The initial studies were performed in healthy men who
were non-obese, non-diabetic, and on high-fat, eucaloric
diets in the absence of potential lipogenic stimulators such
O method for measuring biosynthesis of lipids
(1) Strategy: Avoid the problem of compartmentalization
(2) Complication: How does
O get incorporated during biosynthesis?
(Exchange with solvent H
(3) Applications: DNL and cholesterogenesis
(4) Unresolved problem: true H:C ratio to use
Figure 1A Schematic model of polymerization biosynthesis. The principles of combinatorial probabilities determine the pattern of
containing subunits in newly synthesized polymers. In this simulation, natural abundance or 10% isotopically labelled precursor pools of a subunit ([1-
acetate) combine into a polymer of 8 subunits (palmitate). The population of each of these pools will contain a characteristic distribution of M0, M1, M2,
etc. molecular species (mass isotopomers). These proportions can be represented as a frequency histogram of the mass isotopomer pattern in the polymer
and can be measured by mass spectrometry. After correction for natural abundance, the degree of enrichment of the precursor pool can be calculated by
comparing measured patterns of mass isotopomer abundances to those predicted from theoretical precursor pool enrichments.
Figure 1B Simple numerical example of MIDA principle. The combinatorial probabilities of unlabeled, single-labeled and double-labeled molecular
species are shown for a hypothetical polymer composed of two subunits, with precursor pool labeling either 50% or 10%.
De novo lipogenesis in humans
as fructose or ethanol. All of the factors just noted have
since been tested in humans, that is, the effects of gender,
obesity, diabetes, medical illness, ethanol, fructose, low-fat
diet and overfeeding. These results will now be reviewed.
DNL in women
Women have a different body composition than men, with
larger percent body fat. Faix et al (1993) asked whether this
might be due to a higher rate of DNL under ad libitum diet
conditions. Young women were studied in the luteal and
follicular menstrual phases. In the luteal phase, DNL was
identical in these women as in healthy men (< 1±2g=d). In
the follicular phase, DNL was roughly three times elevated,
and correlated to some extent with estradiol levels. The
absolute rate of hepatic DNL could account for a rather
small body fat burden (< 750 g fat per year), however, in
comparison to dietary fat intake (> 35 kg fat per year).
Although the signal (endocrine or dietary) responsible for
higher DNL in the follicular menstrual phase is an inter-
esting question, differences in DNL are unlikely to explain
body composition differences between the genders.
DNL in obesity and diabetes
Obese, insulin resistant men exhibit modestly increased
fractional DNL (Faix et al, 1993; Schwarz et al, 1993)
while obese, normoinsulinemic men have normal DNL.
The absolute rate of DNL could only account for about four
extra grams of fat=day, however, which is not enough to
account for body fat accrual. More likely, hyperinsulinemia
stimulates hepatic DNL modestly. Consistent with this
explanation, administration of insulin to poorly controlled
diabetics causes an increase in DNL (Christiansen M,
Neese R, Hellerstein M, unpublished observations).
DNL in in¯ammatory diseases
An interesting situation in which DNL is at least to some
extent increased is during infection or in¯ammation, or in
response to recombinant cytokines (such as interleukin-1 or
-6, tumor necrosis factor, etc.) (Grunfeld & Feingold, 1992;
Blackham et al, 1992; Hellerstein et al, 1993). These
settings are also characterized by hypertriglyceridemia
and reduced clearance of TG from plasma (Grunfeld &
Feingold, 1992; Blackham et al, 1992; Hellerstein et al,
1993). Cytokine stimulation of DNL occurs only when
there is liver glycogen present; prior fasting abolishes the
stimulatory effect of tumor necrosis factor on hepatic DNL
in rats (Blackham et al, 1992) and fed-state DNL is more
elevated in AIDS patients than fasted-state DNL (Heller-
stein et al, 1993). The stimulation of DNL by cytokines
does not involve covalent modi®cation of hepatic pyruvate
dehydrogenase (PDH) or inhibition of PDH kinase activity
(Blackham et al, 1992), therefore presumably re¯ects sub-
strate-mediated activation of ¯ux through PDH. Some
workers have speculated that DNL is elevated as part of
an adaptive hyperlipidemic response, which serves as a
primitive host defense mechanism against infection (Grun-
feld & Feingold, 1992). Nevertheless, from a quantitative
point-of-view, total DNL again only represents a few grams
of fat synthesized per day in illnesses such as AIDS
(Hellerstein et al, 1993); body composition changes present
in these conditions cannot be attributed to DNL.
DNL after ethanol (EtOH) ingestion
Siler et al (1996) measured hepatic DNL after administra-
tion of EtOH to normal men (24 ± 48 g EtOH, equivalent to
2 ± 4 alcoholic drinks). The fraction of circulating VLDL-
palmitate contributed by DNL increased considerably,
whether the EtOH was taken alone after an overnight fast
or with a small meal (from < 4% to ca. 30%). Absolute
DNL was still < 2 g after the 48 g EtOH dose, however, or
< 5% of the administered EtOH dose. The great majority of
EtOH was released from the liver as free acetate; 80% of
EtOH clearance appeared in plasma as acetate. Important
secondary consequences of increased plasma acetate ¯ux
were inhibition of FFA release into the circulation (Crouse
et al, 1968) and reduction of whole body fat oxidation. A
model of the fate and consequences of EtOH ingestion in
normal humans is shown schematically (Figure 2) It is of
interest to consider the metabolic consequences that
resulted from DNL being only a minor quantitative sink
for the 2-carbon load on the liver. Because DNL does not
provide a signi®cant disposal route, the liver releases free
acetate, which has priority for oxidation by peripheral
tissues, displaces FFA as a fuel and inhibits further release
of FFA from adipose tissue. Thus, the liver converts EtOH
to a substrate targeted to peripheral tissues' oxidative
needs, rather than directly storing the 2-carbon load as
fat. A similar model can be proposed for surplus CHO in
the liver (see below).
DNL after intake of other potentially lipogenic substrates
(fructose, medium chain TG)
Assimilation of fructose is almost exclusively by the liver
and involves cleavage to triose-phosphates. In addition,
high fructose diets can cause hypertriglyceridemia. For
these reasons, fructose has been considered a potent stimu-
lator of hepatic lipogenesis. Park et al (1992) quanti®ed
hepatic DNL in rats fed 70% fructose diets, using MIDA.
Although DNL was somewhat increased and more hepatic
PDH was in the active form due to inhibition of hepatic
PDH kinase, only 15% of VLDL-palmitate came from
DNL. The hypertriglyceridemia observed could not there-
fore be attributed to increased synthesis of FA from
fructose. Schwarz et al (1993) tested the acute effects of
a fructose load on hepatic DNL in human subjects, using
MIDA. Compared to an equicaloric glucose load (each
given at 10 mg=kg lean body mass=min, by the oral route
for 6 h), fructose caused a qualitative induction of DNL,
that is, the fraction of circulating palmitate derived from
DNL rose to ca. 30%, but < 5% of the fructose load was
converted to VLDL-FA. Although essentially 100% of a
fructose load passes through the hepatic triose-phosphate
pool (Neese et al, 1995), other fates of dietary fructose (for
example, conversion to glucose or glycogen, release as
Figure 2 Key pathways of lipid metabolism in response to acute
consumption of EtOH. Numbers shown re¯ect quantitative response to
EtOH in normal human subjects (Siler et al, 1996).
De novo lipogenesis in humans
lactate) must have priority over conversion to fat. The
lipogenic effects of medium-chain triglycerides are sugges-
tive (Hill et al, 1990), but have not been studied quantita-
tively in animal models or in humans.
DNL on low-fat, high-carbohydrate eucaloric diets
It is important to distinguish between high-CHO (low-fat)
eucaloric diets and surplus CHO hypercaloric diets. The
former refers to the proportion of dietary energy repre-
sented by CHO, under conditions of neutral energy balance;
the latter refers to the absolute quantity of CHO energy
relative to total energy needs.
Parks et al (1998) and Hudgins et al (1996) have
recently studied the effects of eucaloric, low fat:high
CHO diets (15% fat: 70% CHO) on hepatic DNL in
normal human subjects. Some interesting results have
emerged. Either liquid or solid food high CHO diets that
were composed predominantly of simple sugars (60% of
CHO) induce marked fasting hypertriglyceridemia and a
substantial increase in fractional DNL in the post-absorp-
tive state. Solid food diets composed predominantly of
complex CHO (60% of the CHO) also increase fasting
serum triglycerides but do not increase fractional DNL
(Hudgins et al, 1998). VLDL-triglycerides were documen-
ted to come almost 100% from plasma FFA on the complex
CHO low-fat diets. Therefore, there may be `lipogenic' and
`non-lipogenic' high CHO diets. Even the former diets,
however, did not result in a high absolute rate of DNL (less
than 10 g=day, by our calculations, Hudgins et al, 1996).
Taken together, these results suggest that hepatic DNL may
be involved in a regulatory or signaling manner in CHO-
induced hypertriglyceridemia but that the absolute conver-
sion rate of CHO to fat remains rather low on these
DNL during surplus CHO diets
Direct measurements therefore support the conclusion that
hepatic DNL is generally not a quantitatively important
pathway under normal dietary conditions in industrialized
societies. But the question with which we began this
review Ð to what extent is surplus CHO energy converted
to fat through the DNL pathway? Ð is not addressed by the
studies discussed so far. The true test of DNL (Table 1)
requires a surplus of CHO energy intake; that is, CHO
taken in excess of usual CHO oxidative needs in context of
a total dietary energy intake greater than total energy
It should be emphasized that, a priori, conversion to fat
for storage as TG is not the only possible fate of surplus
CHO energy, as just de®ned. `Stoichiometric' arguments
that extra CHO must be converted to fat are not valid
(Table 5), unless CHO energy intake is by itself greater
than TEE (minus protein intake) and occurs for long
enough duration to ®ll whole body glycogen stores to
their maximal capacity. Under other less extreme dietary
conditions that nevertheless represent CHO overfeeding,
the surplus CHO can be stored as glycogen or can replace
fat in the whole-body fuel mixture (Table 5).
A number of overfeeding studies have been performed
in human subjects, providing surplus energy of various
degrees, durations and compositions, while measuring
hepatic DNL. It is useful to divide studies into those in
which CHO energy was not greater than TEE and those in
which it was.
Total CHO energy not in excess of TEE. In the ®rst
category, Neese et al (1994) and Schwarz et al (1995)
measured hepatic DNL by MIDA during ad libitum oral
overfeeding and controlled oral overfeeding, respectively,
in subjects con®ned to a metabolic ward. Subjects in the
former study self-selected an ad libitum, energy surplus diet
for 7 ± 14 d and took in close to 700 g CHO and
4500 kcal=d. The fractional contribution from DNL to
VLDL-FA rose considerably (to 20 ± 30% for palmitate)
but represented only an estimated 1 g fat synthesized per
day. Whole body NP RQ generally was not greater than 1.0
(or only slightly so) but stayed in the 0.95 ± 1.02 range. In a
follow-up study, Schwarz et al (1995) controlled dietary
intake, comparing surplus or de®cient CHO intake for 5 d
periods (at 25%, 50%, 725% and 750% of baseline
energy intake) to eucaloric and to surplus fat (50%) diets.
Fractional DNL (the fraction of VLDL-FA derived from
DNL) and absolute DNL rates were measured. A close
relationship between recent dietary CHO energy and frac-
tional DNL was observed (Figure 3). Indeed, measurement
of fractional DNL was able to correctly identify almost
everyone's recent dietary CHO intake. Stimulation of
fractional DNL was speci®c for dietary CHO surplus: the
50% fat diets showed no effect on DNL (Figure 3).
Nevertheless, absolute hepatic DNL even on the 50%
CHO diet remained quantitatively insigni®cant (Table 6),
representing ca. 3.3 g fat synthesized per day or 9.3 g=d
CHO converted into fat, out of an added dietary CHO
intake in the range of 300 ± 400 g=d.
What was the fate of the added CHO energy, then, if not
conversion to fat? The answer is apparent from Table 7:
whole-body fuel selection adapted to dietary CHO energy.
Therefore, even after an overnight fast, NP RQ was
0.95 0.01 on 50% CHO diets, indicating replacement
of most fat in the fuel mixture by CHO, despite being in the
fasted state. Only 0.5 mg fat were oxidized per hour after an
overnight fast (reduced from 3.6 g fat=h under eucaloric
Table 5 Possible fates of surplus dietary CHO
(1) Storage as glycogen (liver, muscle)
(2) Conversion to fat (DNL in liver, adipose)
(3) Oxidation (replacement of other fuels, i.e. fat)
Figure 3 Hepatic DNL in fasted and fed states, in response to short-term
alterations in dietary energy content. Fractional contribution from DNL to
palmitate isolated from circulating VLDL-fatty acids was determined by
MIDA (Hellerstein et al, 1991; Hellerstein & Neese, 1992). Values not
sharing a common superscript are signi®cantly different (from Schwarz et
De novo lipogenesis in humans
conditions and 5.4 g fat=h during 750% CHO diet).
Associated with the change in fuel selection was a suppres-
sion of adipose lipolysis, signi®cantly increased serum
insulin concentrations, reduced percent of lipolytic ¯ux
oxidized, and perhaps most importantly, increased hepatic
glucose production (HGP, Figure 4). The difference in HGP
between ®ve days of 50% CHO and 7 50% CHO energy
intake was more than 40% (from 1.76 0.09 to
2.48 0.13 mg=kg=min), or roughly the difference between
normal and diabetic subjects (Hellerstein, 1995). Hyper-
glycemia did not result, however, in this setting of
CHO overfeeding. Instead a high-¯ux (glucose-producing=
glucose-oxidizing), normoglycemic fuel economy was
A schematic model of the metabolic response to surplus
CHO energy not in excess of TEE can be proposed from
these results (Figure 5). The key feature of this hypothe-
sized response is that a signal of CHO stores (perhaps liver
glycogen) results in increased HGP, which secondarily
increases fasting insulin secretion; the combination of
elevated glucose plus insulin alters fuel availability and
fuel selection by peripheral tissues. Stimulation of DNL is
observed, but does not play a major quantitative role in the
altered macronutrient ¯uxes, other than perhaps a regula-
tory role in reducing tissue FA oxidation (McGarry &
Total CHO energy in excess of TEE. An important recent
study in the second category was reported by Aarsland et al
(1996). They administered iv plus nasogastric glucose at
rates markedly above TEE for 10 d. Hepatic DNL (frac-
tional and absolute) was measured by MIDA during a 10 h
infusion of [
C]-acetate. After 4 ± 7 d of glucose over-
feeding, hepatic DNL was stimulated 10-fold from baseline
but remained < 1=30 of the whole body value calculated by
What might explain this surprising result, wherein net
whole-body DNL was greater than unidirectional hepatic
DNL? Errors in the precursor pool estimate by MIDA are
not a plausible explanation: hepatic acetyl-CoA enrich-
ments would have to be systematically overestimated by
30-fold (that is, true precursor enrichments of 0.3% instead
of the 9.0% calculated) and from a technical point-of-view,
the mass spectrometric analyses are more reliable at the
high enrichments in these studies than under conditions of
lower DNL. The most reasonable interpretation of these
results is that iv glucose overfeeding induces DNL in a
`hidden compartment'. Aarsland et al (1996) concluded
that adipose DNL must be occurring. This may be true, or it
is also possible that the 10 h isotope infusion was of
insuf®cient duration to allow export of newly synthesized
TG from the hepatic cytosolic TG pool, if the hepatic TG
pool is expanded under these conditions. Neither possibility
has been evaluated experimentally.
There have been no studies reported, to date, of even
longer periods of CHO overfeeding at rates in excess of
TEE (that is, comparable to the Guru Walla model, Pasquet
et al, 1992) with direct measurement of liver or adipose
Are surplus CHO calories not in excess of TEE therefore
`free' of risk for adding to body fat stores? It is important
to be clear about the effects of surplus CHO energy not in
excess of TEE on whole body fat balances, if DNL is not a
quantitatively important pathway. This has been widely
misinterpreted in the lay literature. Although surplus CHO
energy at these levels may not be converted directly to fat
Figure 4 Fasting hepatic glucose production (Ra glucose) in response to
short-term alterations in dietary energy content. Values shown are percent
change from baseline (eucaloric) value, in mg=kg=min (from Schwarz et
Figure 5 Hypothetical model of metabolic mechanisms responsible for
sensitivity of whole-body fuel selection to dietary CHO intake. All
precesses other than the box labeled `?' were measured (Schwarz et al,
1995). INS, insulin; ox, oxidation.
Table 6 Absolute DNL (50% CHO diet, n 6)
Fractional DNL Ks T
[TG] TG Prod. Abs. DNL Glc to fat
25.4 3.1 0.370 0.059 2.21 0.44 104 7 28.0 5.0 3.3 0.8 9.3 2.3
Fractional DNL, fraction of palmitate in VLDL derived from DNL; Ks, rate constant of VLDL-TG turnover;
abs. DNL, absolute rate of DNL; Glc to fat, absolute rate of glucose conversion to fat (from Schwarz et al, 1995).
Table 7 Whole body fuel utilization on different diets (NPRQ)
0.84 0.01 0.95 0.01 0.77 0.01 0.91 0.01 0.80 0.01 0.84 0.02
De novo lipogenesis in humans
in large quantities, CHO replaces fat as fuel by the body;
total fat oxidation can be almost completely turned off by
intake of surplus CHO (Table 7). Body fat is thereby spared
by surplus CHO; despite the absence of DNL, body fat can
accumulate, as long as there is any fat in the diet.
Metabolic response to surplus CHO energy intake in
humans: the emerging model
Although a number of questions remain unresolved con-
cerning the threshold for and tissue site of DNL in humans
in response to long-term surplus CHO energy, the results to
date allow a model to be proposed (Table 8, Figure 5).
Several points deserve emphasis:
1. DNL is not the `pathway of ®rst resort' in response to
intake of surplus CHO. The ®rst and quantitatively most
important responses to increased CHO intake are increased
glycogen storage and increased whole body CHO oxida-
tion, not increased CHO conversion to fat. Although net fat
balance and cumulative indirect calorimetry can not distin-
guish between direct CHO oxidation and lipogenesis=fat
oxidation over the course of 24 h, and RQ must equal FQ
under conditions of energy balance whether or not DNL
occurs (as discussed by Flatt, 1987), there are extremely
important functional consequences of the particular orga-
nization that appears to have evolved. Alternative systems
could exist: there could have been a circadian cycle for
lipogenesis and fat oxidation, for example (Figure 6), with
storage of CHO as fat during the daytime then mobilization
and oxidation of fat at night. The net result might be the
same under conditions of energy balance, but the speci®c
pathways involved and the possible disorders of their
regulation (diseases) would be quite different. Instead,
CHO energy taken in excess of postprandial energy
requirements is stored in modestly expandable tissue gly-
cogen stores during the daytime and is then released from
glucose producing tissues (the liver) at a rate proportional
to its prior accumulation, for oxidation during the overnight
fast. Rather than observing highly cyclic values of RQ and
a diurnal sequence of fat storage followed by fat oxidation
in the presence of CHO surpluses, cyclic variations in RQ
are damped and the system settles into a stable CHO
oxidizing mode (Figure 6). The diurnal cycle consists of
daytime CHO storage and night-time CHO release, rather
than daytime DNL followed by night-time fat oxidation.
Viewed from this perspective, the macronutrient regula-
tory system appears to be organized to match fuel selection
to recent CHO intake. It has been long recognized (Cahill,
1976) that animals must reduce CHO oxidation in times of
energy starvation, to preserve lean body mass (by reducing
the need for gluconeogeneses from amino acids). The
converse adaptation seems also to apply, namely, that
CHO oxidation is increased in times of CHO surplus.
Moreover, both adaptations involve the liver and modula-
tion of HGP, rather than just direct oxidation of expanded
or reduced CHO stores in peripheral tissues.
2. Many functional consequences can be predicted from
point #1. A number of functional consequences for
human metabolic physiology and disease can be predicted
a priori from the simple notion that CHO is not readily
converted to fat and that CHO stores in the body are
limited, as Flatt has discussed elsewhere (Flatt, 1987).
Many of these predictions have experimental support, as
(i) Whole body fuel selection is controlled by and respon-
sive to recent CHO intake (and CHO stores in the
body). This applies not only under conditions of
macronutrient and energy balance, but also during
periods of surplus (or de®cient) CHO intake, that is,
if CHO stores are changing.
(ii) HGP must be modulated in response to the balance
between dietary CHO and whole body CHO oxida-
tion. Because 24 h CHO oxidation must re¯ect total
CHO intake but food intake is cyclic, glucose release
at night (the interprandial period) must be sensitive to
CHO intake during the day (the prandial period)
(Figure 4). As a result, hepatic insulin sensitivity or
pancreatic insulin secretion must be capable of mod-
ulation and must respond to diet. Without variable
HGP, matching of 24 h fuel selection to dietary CHO
would not work. Therefore, HGP is not a constant,
like serum sodium concentration or osmolality, but
depends upon recent CHO balances. Studies in rats
(Neese et al, 1995), normal humans (Schwarz et al,
1995; Clore et al, 1995) and type II diabetes mellitus
(Wing et al, 1994; Kelley et al, 1993; Christiansen et
al, 1995) have documented the responsiveness of
HGP to recent CHO energy intake. Biochemical
mechanisms for this adaptation have been proposed.
Liver glycogen content affects its own turnover by
increasing sensitivity to glycogen phosphorylase
(Hers, 1976). Therefore, expanded hepatic glycogen
content increases HGP and contracted glycogen con-
tent reduces HGP (Schwarz et al, 1995; Clore et al,
1995; Neese et al, 1995). Insulin secretion has also
been reported to be in¯uenced by body CHO stores
(Lilavivathana et al, 1978). Stated differently, the
need for modulable HGP in response to high dietary
CHO represents a form of physiologic (adaptive,
Table 8 Conclusions: what happens when surplus CHO is consumed?
(1) CHO not converted to fat (mysterious block)
(2) Instead: CHO burnt, fat not; CHO stores ®ll up
(3) Provides a system for matching fuel selection to recent CHO intake
(and perhaps controlling intake)
(4) System wouldn't work if a safety-valve (CHO to fat) were open
Figure 6 Circadian pattern of fuel selection (expressed as NPRQ):
potential models of response to surplus dietary CHO. In model 1, NPRQ
is close to 1.0 and relatively constant over the circadian feeding cycle. In
model 2, NPRQ varies considerably in response to feeding or fasting, with
low NPRQ during the overnight post-absorptive phase and NPRQ > 1.0
during the daytime absorptive phase.
De novo lipogenesis in humans
programmed) hepatic insulin resistance. The implica-
tion for diseases related to insulin resistance, includ-
ing type II diabetes, have not been fully explored,
although this metabolic adaptation appears to repre-
sent the physiologic basis of the therapeutic response
to energy restriction (`weight loss') in type II diabetes
(Wing et al, 1994; Kelley et al, 1993; Christiansen et
(iii) The liver is therefore responsible for whole-body fuel
selection in the post-absorptive state. The rate of
release of glucose into the circulation, and the sec-
ondary consequences of this, particularly the altera-
tions in insulin secretion, profoundly affect the fuel
mixture selected by tissues. Implications for the role
of the liver in the etiology of obesity have only
recently been considered. Pagliassotti et al (1997a,b)
have shown that non-suppressability of HGP by insu-
lin is the best predictor of obesity in rats fed ad libitum
high-fat diets. Over-expression of the GNG enzyme
PEP-CK in the liver of transgenic mice (Valera et al,
1994) and rats (Rosella et al, 1995) results in hyper-
insulinemia, obesity and ultimately hyperglycemia.
Therefore, a liver enzyme can be responsible for
obesity and its sequelae. It is of interest to note that
hepatic insulin can be predicted to exert opposite
effects on body composition than muscle insulin
resistance exerts (Swinburn et al, 1991). The former
is characterized by increased HGP and peripheral
glucose utilization (a high-¯ux state for glucose),
with reduced fat oxidation and increased CHO oxida-
tion in the whole body; the latter by reduced glucose
utilization and production (a low-¯ux state for glu-
cose), with reduced CHO oxidation and increased fat
oxidation in the whole body (Swinburn et al, 1991).
Accordingly, the presence of hepatic insulin resistance
out of proportion to muscle insulin resistance will tend
to increase body fat stores, whereas muscle insulin
resistance disproportionate to liver will tend to reduce
In this context, it is also apparent that that the liver is
more likely to contribute to obesity by overproducing
glucose than by converting CHO to fat. Though counter-
intuitive on the surface, this conclusion follows from the
basic physiologic premises just discussed and has impor-
tant implications for genetic and pathophysiologic
investigations of the etiology of obesity.
(iv) Finally, the absence of an analogous mechanism for
matching fuel selection to changes in fat intake should
be noted (Figure 7). Thus, low fat diets per se do not
cause changes in HGP or help drive RQ to match FQ;
instead, it is the reciprocal change in CHO intake that
appears to provide the signals that alter fuel selection,
including tissue oxidation of fat. Sidossis & Wolfe
(1996) have recently termed this `the Randle cycle
reversed', namely, that CHO availability controls
tissue fat oxidation more than FFA availability con-
trols tissue CHO oxidation, if both substrates are
present concurrently. Metabolic mechanisms to
explain regulation of tissue fat oxidation by CHO
availability are well described (McGarry & Foster,
1980). The absence of speci®c mechanisms to match
fat oxidation to increased fat intake (Figure 7) con-
tribute to sensitivity to dietary fat-induced obesity,
(Pagliassotti et al, 1997a,b).
Unanswered questions concerning the regulation and
functions of DNL
There are still many unanswered questions concerning the
regulation of DNL and its role in human physiology or
(1) What constrainsDNLunderusual dietaryconditions? Is
there a signal that restrains DNL in liver or adipose tissue,
despite the availability of substrate, enzymes and co-
factors and the inducibility of lipogenic gene expression?
Studies in animals and in vitro suggest that speci®c FAs
exert negative feedback on expression of DNL genes and
enzymes (Girard et al, 1994; Sul et al, 1993). Translation
of these ®ndings to physiology is not obvious, however.
How much fat, and in what tissue, feeds back on the DNL
Results with a transgenic=knockout mouse model of
dietary fat malabsorption may be informative in this
regard. Young et al have developed a mouse with the
endogenous apoB gene knocked out and a human apoB
transgene expressed in liver only (Young et al, 1995).
Accordingly, these mice are unable to form chylomicrons
or ef®ciently absorb dietary fat; the result is that CHO
represents a high percent of absorbed nutrients. Despite
the marked reduction of dietary fat absorption (estimated
to be 10 ± 20% of normal, based on serum concentrations
of lipid-soluble vitamins), these animals maintain normal
plasma concentration of triglycerides and cholesterol and
are able to accumulate body fat, though at a much lower
rate than in normal mice (Young et al, 1995). These
observations suggested that DNL might occur at a very
high rate, to maintain hepatic lipoprotein secretion and
adipose fat deposition. To address this question, we have
measured DNL, FFA ¯ux and re-esteri®cation to hepatic
TG in these mice (Turner et al, 1998). The most striking
result was a greatly accelerated incorporation of plasma
C-FFA into hepatic and VLDL-TG (that is, accelerated
re-esteri®cation of plasma FFA). Hepatic fractional DNL
was only slightly elevated (15% vs 5%) while adipose
DNL appeared to be identical or even lower after several
weeks of oral
C-acetate feeding in chylomicron-de®cient
mice compared to normal mice.
We (Turner et al, 1998) concluded from these results
that very low dietary fat absorption does not induce DNL
in mice; and that hepatic DNL may be inhibited by the
presence of FA in adipose stores even in small quantities,
Figure 7 Proposed metabolic response to surplus dietary CHO or fat.
Effects of increased CHO intake are shown from the left; effects of increased
fat intake from the right. Abbreviations are as in the text. Inhib, inhibited.
De novo lipogenesis in humans
by virtue of the capacity of the liver to avidly take up and
re-esterify adipose-derived FFA. Thus, FA in adipose
stores may inhibit DNL in liver. If true, a prediction is
that lipolysis inhibitors might markedly disinhibit hepatic
DNL in these mice or on low-fat diets in general.
Another approach to the question of factors limiting DNL
is by comparison of different substrates. The liver may
simply have a limited capacity for oxidation of pyruvate
derived from glycolysis to acetyl-CoA. Results with EtOH
(Figure 2) do not support this explanation, however (Siler et
al, 1996): although EtOH enters the liver at the 2-carbon
level, the quantitatively modest hepatic lipogenesis and the
release of large amounts of free acetate after EtOH ingestion
suggests either limited capacity to convert acetate to acetyl-
CoA or a block in conversion of acetyl-CoA to fat (that is,
intrinsically limited DNL capacity), with subsequent acetyl-
What is the contribution of adipose DNL in humans?
Although most work has focused on hepatic DNL in
humans and enzyme levels have suggested adipose to be
a weakly lipogenic tissue in people (Shrago et al, 1971;
Weiss et al, 1986; Sjostrom, 1973), recent evidence (Aars-
land et al, 1996) suggests that adipose tissue may be the
place to look for quantitatively signi®cant DNL during
overfeeding, particularly by the parenteral route. Before
adipose DNL can be measured in vivo, however, two
technical problems will need to be solved. Firstly, the
huge adipose TG pool size results in such large dilution
of newly synthesized TG that reliable measurement of label
incorporation has not been possible in human subjects. One
solution might be to carry out much longer isotopic studies;
another idea is to isolate diglycerides and=or monoglycer-
ides from adipose (since these are believed to be either en
route to or from TG). The second problem is how to
exclude the possible contribution from hepatic-synthesized
TG that were transported to adipose tissue. Neither techni-
cal problem has yet been solved.
It should be pointed out here that quantitatively signi®-
cant hepatic DNL is not a physiologic impossibility. If
measurements in vivo under lipogenic conditions showed
the following values, quantitatively signi®cant DNL would
be present: fractional DNL contribution 75 ± 80% of non-
essential FA in VLDL-TG, non-essential FA comprising
75 ± 80% of TG, and total VLDL-TG production and
clearance rates 4- or 5-fold elevated (to 100 ± 125 g=d).
Hepatic DNL would then contribute 60 ± 75 g new fat
synthesis daily. Hepatic DNL rates of this magnitude
would have a major impact on macronutrient balances,
but have not been found.
(3) Are there any inborn errors of metabolism leading to
elevated rates of DNL? At least two genetic conditions
might be associated with greatly increased rates of DNL:
one found in nature, one induced by recombinant DNA
techniques. Glycogen storage disease-type I (GSD-1) is a
relatively common inborn error of metabolism due to the
absence of hepatic glucose-6-phosphatase activity (Chen &
Burchell, 1994). Hepatic glycogenosis and fat accumula-
tion, elevated plasma lactate concentrations, marked hyper-
triglyceridemia, absence of serum ketones and border-
line-hypoglycemia are included in the phenotype
of GSD-1. A role for hepatic DNL in the TG
accumulation in liver and plasma has been proposed and
might be predicted from the extremely high availability of
3-carbon (lactate=pyruvate) and glucose (glycogen) precur-
sors in the liver. DNL has not yet been directly measured in
Shimano et al (1996) over-expressed the steroid response
element-binding protein (SRE-BP) in transgenic mice.
SRE-BP stimulates transcription and expression of choles-
terogenic and DNL genes. Mice over-expressing SRE-BP
exhibited markedly elevated serum TG and cholesterol
concentrations and fatty livers, with evidence of increased
DNL (Shimano et al, 1996), although the quantitative
contribution from DNL was not assessed and the animals
did not become obese. Whether an analogous genetic
disorder exists in humans has not been determined.
(4) What about embryonic (fetal) and=or mammary gland
DNL? One possible explanation for why the enzymatic
machinery for DNL is present but used so sparingly in adult
humans is that this represents a vestigial pathway in post-
natal life, that is, the predominant function of DNL occurs
in utero. The rationale behind this hypothesis is that during
the third trimester in particular the human fetus has a very
large demand for lipids, for deposition of subcutaneous fat
as well as for myelination of the developing central nervous
system. The brain is 60% fat by wet weight and represents
20% of body weight in the newborn human (360 g fat in the
brain of a 3 kg infant). If body fat is 20% of body weight, in
addition (600 g) close to 1 kg FA needs to be deposited,
mainly during the third trimester. The fetus therefore needs
upwards of 10 ± 12 g FA=d for storage purposes alone.
Transport of lipoproteins or FFA across the placenta from
the mother is inef®cient, however, and appears inadequate
to meet the lipid requirements of the developing fetus. A
role for greatly elevated rates of fetal lipogenesis might
therefore be predicted. Fetal DNL has not been directly
measured in humans. Some indirect evidence for high DNL
in utero comes from studies of premature infants, who often
exhibit extraordinarily high values of NP RQ (for example,
> 1.20, Dr. W. Chwals, personal communication) in
response to parenteral nutrition even with intralipid (trigly-
ceride emulsions) included. If these premature infants are
still manifesting in utero physiology, it might be inferred
that DNL is indeed elevated in utero. This hypothesis
Milk secreted by the mammary gland is high in fat.
Although the source of this fat in human breast milk has not
been directly measured, lactating mammary gland
expresses DNL enzymes at high levels. Mammary gland
DNL represents a potential role for this pathway, but this
has not been yet proven.
(5) Are there other regulatory functions of DNL in tis-
sues? The ®rst committed step in the lipogenic pathway is
synthesis of malonyl-CoA from acetyl-CoA and carbon
dioxide, catalyzed by the enzyme acetyl-CoA carboxylase
(Bloch, 1977; Girard et al, 1994). Malonyl-CoA inhibits
transport of FA-CoAs into mitochondria by inhibiting
carnitine acyl-transferase I activity (McGarry & Foster,
1980). Therefore, within a given cell, the occurrence of
lipogenesis is believed generally to be inconsistent with b-
oxidation of fatty acids and to promote cytosolic esteri®ca-
tion of FA-CoAs.
De novo lipogenesis in humans
This indirect function of the lipogenesis pathway has
been proposed to play a regulatory role in several tissues. In
liver, malonyl-CoA has anti-ketogenic and pro-esterifying
effects on FA-CoAs, thereby promoting TG synthesis and
assembly of ApoB containing particles. In muscle, mal-
onyl-CoA may serve as an important signal regulating fuel
selection (despite the lack of recognized lipogenic fate for
malonyl-CoA in muscle) (Ruderman & Herrera, 1968;
Ruderman et al, 1969; Ruderman & Haudenschild, 1984).
In pancreas, Prentki et al (1992) have proposed that
malonyl-CoA levels increase in response to b-cell glucose
utilization, leading to an accumulation of FA-CoAs in the
cell cytosol, which in turn serves as a metabolic signal for
insulin secretion. Components of the DNL pathway may
therefore have important regulatory functions in tissues that
are not considered to be classically lipogenic.
Another possible regulatory function of DNL is the
synthesis of fatty acids, such as myristate, that are used
for acylation of proteins (Johnson et al, 1994). Whether the
myristate used for myristoylation is derived from diet or
DNL is unknown, at present. If the source is diet, then
myristate should be considered an essential FA. More
likely, these FAs are synthesized de novo from acetyl-
CoA, in which case DNL might be regulated for reasons
other than the needs of the energy economy alone.
(6) What would be the physiologic consequences if DNL
were a more active pathway? It is interesting to speculate
what would happen metabolically if DNL in fact served as
the disposal route for transient surpluses of CHO. One
physiologic model of this situation, wherein the DNL
pathway is wide-open might be the marmot prior to
hibernation (Young & Sims, 1977). These animals deposit
large quantities of body fat in preparation for hibernation
and DNL has been reported to be markedly elevated. Might
DNL play an important role in this setting of (adaptive)
positive fat balance? According to the model of Flatt
(1987), carbohydrate stores are postulated to in¯uence
food intake and satiety. Evidence has been presented in
support of a role for liver glycogen (or other metabolites
related to hepatic hexoses) in appetite regulation (Sullivan
et al, 1974; Russek, 1963, 1981; Russek & Stevensen,
1972), although the hypothesis remains largely unproven.
The occurrence of a rapid ¯ux through DNL from CHO in
pre-hibernation animals might keep tissue glycogen stores
relatively low, prevent satiety and promote continued food
intake, in addition to directly contributing fat to body
stores. It should be noted that tissue fuel selection is not
matched to food intake under these conditions (by intent),
unlike during most of an adult animal's life.
If pre-hibernating mammals indeed exhibit wide-open
¯ow through the DNL pathway, some predictions (see
above) might be testable. One need not postulate increased
HGP or reduced tissue fat oxidation in the post-absorptive
state in these animals, for example, despite a marked excess
of CHO intake, if surplus CHO is removed from tissue
stores by conversion to fat during the absorptive phase.
Another way to test the functional consequences of remov-
ing normally constrained DNL might be through
transgenic techniques (that is, over-expressing the enzymes
of DNL and=or pyruvate oxidation in liver or adipose
(7) What are the physiologic functions of DNL? This
discussion raises the larger question: What are the physio-
logic functions of DNL? A number of possibilities can be
considered (Table 9). Most of these have been discussed
above, but no de®nitive answer to the question can be
provided at this time. Differentiating between these possi-
bilities will have important implications for a number of
questions in metabolic regulation and the control of energy
(8) Does DNL contribute to human disease? Even with
relatively low quantitative ¯uxes through the pathway,
DNL may be involved in a variety of diseases (Table 10).
FA synthesized through DNL in the liver appear to be
preferentially secreted as VLDL-TG rather than stored in
the cytosolic TG pool (Gibbons, 1990; Gibbons & Burn-
ham, 1991). If hepatic FA from DNL are more effective at
protecting apoB from proteolysis and promoting VLDL-
particle assembly than FA derived from plasma FFA or
lipoprotein remnants (Gibbons, 1990; Gibbons & Burnham,
1991), then hyperlipidemia associated with stimulators of
DNL (EtOH, fructose, cytokines) might re¯ect effects on
lipoprotein assembly out of proportion to its quantitative
contribution of new FAs. Fatty liver, the precursor to
®brosis in alcoholic cirrhosis, might also be related to
DNL (Siler et al, 1996). Hepatic fat deposition in GSD-1
may contribute to the tendency for liver tumors (Chen &
Burchell, 1994) and, as discussed above, these FA could
derive in part from the DNL pathway. Another result of
elevated rates of DNL is a change in FA composition of
VLDL-TG to a more saturated pro®le (Hudgins et al,
1996). If membrane FA composition in certain cells
throughout the body also become more saturated as a
result, a variety of physiologic consequences might result,
including insulin resistance (Storlein et al, 1996). Finally,
alterations in intracellular signaling pathways related to
changes in myristolylation, palmitoylation, malonyl-CoA
availability, etc. (McGarry & Foster, 1980; Prentki et al,
1992; Johnson et al, 1994) could be involved in disorders
Table 9 What are the physiologic functions of de novo lipogenesis? Ð
(1) There are no functions (vestigial pathway in humans)
(2) Necessary in embryonic development Ð for CNS lipid synthesis
(vestigial pathway in post-natal life)
(3) Important on low-fat diets Ð suppressed by un-natural modern diet
(vestigial pathway culturally)
(4) Only important after long-term surplus CHO (glycogen over¯ow
(5) Serves signal or regulatory functions (e.g. antiketogenesis, tissue fat
oxidation, insulin synthesis, etc.)
Table 10 Possible contribution of hepatic lipid synthesis to human
(I) Plasma lipids and lipoproteins (atherogenesis)
(II) Energy balance (obesity, insulin resistance)
(III) Bile synthesis (cholelithiasis, atherosclerosis)
(IV) Hepatic fat (fatty liver, cirrhosis)
(V) Cell membrane FA composition (physiologic consequences)
(VI) Intracellular signaling (myristoylation, malonyl-CoA, etc.)
De novo lipogenesis in humans
related to these mediators. These potential roles of DNL in
disease are speculative at present, but deserve investigation.
DNL is not the pathway of ®rst resort for added dietary CHO in
humans, at least on Western (high-fat) diets. DNL can occur,
but it generally does not. A `functional block' therefore exists
between CHO and fat in humans, analogous to the absolute
biochemical block in the direction from fat to carbohydrate in
all animals. Therefore, the two major macronutrient energy
sources are not interconvertible currencies in the mammalian
organism; they must be considered separately and are prob-
ably regulated independently, by separate signals and toward
separate ends. The major insight concerning DNL is therefore
a negative one. Many questions related to this central observa-
tion still remain unanswered: what is the functional signi®-
cance of DNL in adult life (Table 9)? What are the ultimate
limits of DNL in humans? Is DNL only used as a ®nal `safety-
valve' for CHO in the organism? What constrains DNL in
human lipogenic tissues? Are there regulatory roles played by
DNL that we have not yet identi®ed? Regardless of the
answers to these questions, the metabolic and clinical con-
sequences of the apparent functional block between CHO and
fat are profound and have only begun to be understood.
Aarsland A, Chinkes D & Wolfe RR (1996): Contributions of de novo
synthesis of fatty acids to total VLDL-triglyceride secretion during
prolonged hyperglycemia=hyperinsulinemia in normal man. J. Clin.
Invest. 98, 2008 ± 2017.
Acheson K, Schutz T, Bessard K, Anatharaman AS, Flatt J-P, et al (1988):
Glycogen storage capacity and de novo lipogenesis during massive
carbohydrate overfeeding in man. Am. J. Clin. Nutr. 48, 240 ± 247.
Acheson KJ, Flatt J-P & Jequier E (1982): Glycogen synthesis versus
lipogenesis after a 500-g carbohydrate meal. Metabolism 31,
1234 ± 1240.
Acheson KJ, Schutz Y, Bessard T, Ravussin E, Jequier E et al (1984):
Nutritional in¯uences on lipogenesis and thermogenesis after a carbo-
hydrate meal. Am. J. Physiol. 246, E62 ± E70.
Askanazi O, Rosenbaum SH, Hyman AJ, Silverberg PA, Milic-Emili J, et
al (1980): Respiratory changes induced by large glucose loads of total
parenteral nutrition. JAMA 243, 1444 ± 1447.
Blackham M, Cesar D, Park O-J, Wu K, Kaempfer S, et al (1992): Effects
of recombinant monokines on hepatic pyruvate dehydrogenase (PDH),
PDH kinase, de novo lipogenesis and plasma triglycerides. Abolition by
prior fasting. Biochem. J. 284, 129 ± 135.
Bloch K (1977): Control mechanisms in the synthesis of saturated fatty
acids. Ann. Rev. Biochem. 46, 263 ± 298.
Cahill GF (1976): Starvation in man. Clin. Endocrinol. Metab. 5,
397 ± 415.
Chen Y-T & Burchell A (1994): Glycogen storage diseases. In The
Metabolic Basis of Inherited Diseases, 7th ed, eds. Scriver CR, Beaudet
AL, Sly WS, Valle D. McGraw-Hill Inc, New York, pp 935 ± 965.
Christiansen M, Linfoot P, Neese R, Turner S & Hellerstein M (1995):
Gluconeogenesis (GNG) in NIDDM: effect of energy restriction. Dia-
betes 44, 55A.
Clore J, Helm S & Blackard W (1995): Loss of hepatic autoregulation
after carbohydrate overfeeding in normal men. J. Clin. Invest. 96,
1967 ± 1972.
Crouse JR, Gerson CD, DeCarli LM & Lieber CS (1968): Role of acetate in
the reduction of plasma free fatty acids produced by ethanol in man. J.
Lipid Res. 9, 509 ± 512.
Dietschy JM & Brown MS (1974): Effect of alterations of the speci®c
activity of intracellular acetyl-CoA pool on apparent rates of hepatic
cholesterogenesis. J. Lipid Res. 15, 508 ± 516.
Elia M & Livesey G (1988): Theory and validity of indirect calorimetry
during net lipid synthesis. Am. J. Clin. Nutr. 47, 591 ± 607.
Faix D, Neese RA, Kletke C, Walden S, Cesar D, et al (1993): Quanti®ca-
tion of periodicities in menstrual and diurnal rates of cholesterol and fat
synthesis in humans. J. Lipid Res. 34, 2063 ± 2075.
Flatt JP (1978): The biochemistry of energy expenditure. Rec. Adv. Obesity
Res. 2, 211 ± 217.
Flatt JP (1987): Dietary fat, carbohydrate balance, and weight maintenance:
effects of exercise. Am. J. Clin. Nutr. 45, 296 ± 306.
Florkin M & Stotz EH (1977): Comprehensive Biochemistry. A History of
Biochemistry, vol 32. New York: Elsevier Scienti®c, pp 277 ± 278.
Gibbons G & Burnham F (1991): Effect of nutritional state on the
utilization of fatty acids for hepatic triacylglycerol synthesis and secre-
tion as very-low-density lipoprotein. Biochem. J. 275, 87 ± 92.
Gibbons GF (1990): Assembly and secretion of hepatic very-low-density
lipoprotein. Biochem. J. 268, 1 ± 13.
Girard J, Perdereau D, Foufelle F, Prip-Buus C & Ferre P (1994):
Regulation of lipogenic enzyme gene expression by nutrients and
hormones. FASEB J. 8, 36 ± 42.
Grunfeld C & Feingold K (1992): Metabolic disturbances and wasting in
the acquired immunode®ciency syndrome. N. Engl. J. Med. 327, 329 ±
Hellerstein MK (1991): Relationship between precursor enrichment and
ratio of excess M
isotopomer frequencies in a secreted
polymer. J. Biol. Chem. 266, 10920 ± 10924.
Hellerstein MK (1995): Isotopic studies of carbohydrate metabolism in
non-insulin-dependent diabetes mellitus. Curr. Opin. Endocrinol. Diab.
2, 518 ± 529.
Hellerstein MK, Christiansen M, Kaempfer S, Kletke C, Wu K et al (1991):
Measurement of de novo hepatic lipogenesis in humans using stable
isotopes. J. Clin. Invest. 87, 1841 ± 1852.
Hellerstein MK, Grunfeld C, Wu K, Christiansen M, Kaempfer S,
et al (1993): Increased de novo hepatic lipogenesis in human
immunode®ciency virus infection. J. Clin. Endocrinol. Metab. 76,
559 ± 565.
Hellerstein MK & Neese R (1992): Mass isotopomer distribution analysis:
a technique for measuring biosynthesis and turnover of polymers. Am. J.
Physiol. 263, E988 ± E1001.
Hellerstein MK, Schwarz J-M & Neese RA (1996): Regulation of hepatic
de novo lipogenesis in humans. Ann. Rev. Nutr. 16, 523 ± 557.
Hellerstein MK, Wu K, Kaempfer S, Kletke C & Shackleton CHL (1991):
Sampling the lipogenic hepatic acetyl-CoA pool in vivo in the rat.
Comparison of xenobiotic probe to values predicted from isotopomeric
distribution in circulating lipids and measurement of lipogenesis and
acetyl-CoA dilution. J. Biol. Chem. 266, 10912 ± 10919.
Hill JO, Peters JC, Swift LL, Yang D, Sharp T, Abunrad N & Greene HL
(1990): Changes in blood lipids during six days of overfeeding with
medium or long-chain triglycerides. J. Lipid Res. 31, 407 ± 416.
Hers HG (1976): The control of glycogen metabolism in the liver. Ann.
Rev. Biochem. 45, 167 ± 189.
Hirsch J (1965): Fatty acid patterns in human adipose tissue. In Handbook
of Physiology, eds. JF Cahill & AE Renold. Baltimore: Waverly, pp
181 ± 189.
Hudgins L, Seidman C, Diakun J & Hirsch J (1998): Human fatty acid
synthesis reduced after the substitution of dietary starch for sugar. Am. J.
Clin. Nutr. 67, 631 ± 639.
Hudgins LC, Hellerstein M, Seidman C, Neese R, Diakun J & Hirsch J
(1996): Human fatty acid synthesis is stimulated by a eucaloric low-fat,
high carbohydrate diet. J. Clin. Invest. 97, 2081 ± 2091.
quier E, Acheson K & Schutz Y (1987): Assessment of energy expen-
diture and fuel utilization in man. Ann. Rev. Nutr. 7, 187 ± 208.
Johnson DR, Bhatnager RS, Knoll LJ & Gordon JI (1994): Genetic and
biochemical studies of protein N-myristoylation. Ann. Rev. Biochem. 63,
869 ± 914.
Jones PJ, Namchuk GL & Pederson RA (1995): Meal frequency in¯uences
circulating hormone levels but not lipogenesis rates in humans. Metab.
Clin. Exp. 44, 218 ± 223.
Jungas RL (1968): Fatty acid synthesis in adipose tissue incubated in
tritiated water. Biochemistry 7, 3708 ± 3717.
Katz J (1985): Determination of gluconeogenesis in vivo with
substrates. Am. J. Physiol. 248, R391 ± R399.
Kelley D E, Wing R, Buonocore C, Staris J, Polonsky K & Fitzsimmons M
(1993): Relative effects of calorie restriction and weight loss in non-
insulin-dependent diabetes mellitus. J. Clin. Endocrinol. Metab. 77,
1287 ± 1293.
Lawes JF & Gilbert JH (1886): On the source of fat of the animal body.
Phil. Mag. (4), 439 ± 451.
Lee WNP, Bassilian S, Guo Z, Schoeller D, Edmond DJ, et al (1994):
Measurement of fractional lipid synthesis using deuterated water (
and mass isotopomer analysis. Am. J. Physiol. 266, E372 ± E383.
Leitch CA & Jones PJH (1993): Measurement of human lipogenesis using
deuterium incorporation. J. Lipid Res. 34, 157 ± 163.
Lilavivathana U, Campbell RG & Brodows RG (1978): Control of insulin
during fasting in man. Metabolism 27, 815 ± 821.
McGarry JD & Foster DW (1980): Regulation of hepatic fatty acid oxidation
and ketone body production. Ann. Rev. Biochem. 49, 395 ± 420.
De novo lipogenesis in humans
Neese RA, Benowitz NL, Hoh R, Faix D, LaBua A, et al (1994): Metabolic
interactions between surplus dietary energy intake and cigarette smok-
ing or its cessation. Am. J. Physiol. 267, E1023 ± E1034.
Neese RA, Faix D, Schwarz J-M, Turner SM, Vu D, et al (1995):
Measurement of gluconeogenesis and rate of appearance of intrahepatic
triose-phosphate and its regulation by substrates by mass isotopomer
distribution analysis (MIDA). Testing of assumptions and potential
problems. J. Biol. Chem. 270, 14452 ± 14463.
Pagliassotti M, Gayles E & Hill JO (1997b): Fat and energy balance. In
Lipids and Syndromes of Insulin Resistance. Ann. NY Acad. Sci. 827,
431 ± 448.
Pagliassotti M, Harton T, Gayles E, Koppenhafer T, Rosenzweig T & Hill
JO (1997a): Reduced insulin suppression of glucose appearance is
related to susceptibility to dietary obesity in rats. Am. J. Physiol. 272,
R1264 ± R1270.
Park OJ, Cesar D, Faix D, Wu K, Shackleton CHL & Hellerstein MK
(1992): Mechanisms of fructose-induced hyperglyceridemia in the rat:
activation of hepatic pyruvate dehydrogenase (PDH) through inhibition
of PDH kinase. Biochem. J. 282, 753 ± 757.
Parks E, Christiansen M, Krauss R, Dare D, Neese R & Hellerstein M
(1998): Effects of high carbohydrate diet on very low density-lipopro-
tein (VLDL) particle synthesis and composition in normolipidemic and
hypertriglyceridemic individuals. Submitted.
Pasquet P, Brigant L, Foment A, Koppert G, Bard D, et al (1992): Massive
overfeeding and energy balance in men: the Guru Walla model. Am. J.
Clin. Nutr. 56, 483 ± 490.
Passmore R & Swindells YE (1963): Observations on the respiratory
quotients and weight gain of man after eating large quantities of
carbohydrate. Brit. J. Nutr. 17, 331 ± 339.
Petrek J, Hudgins L, Ho M, Bajorunas D & Hirsch J (1997): Fatty acid
composition of adipose tissue, an indication of dietary fatty acids and
breast cancer prognosis. J. Clin. Oncol. 15, 1377 ± 1384.
Prentki M, Vischer S, Glennon MC, Regazzi R, Deeney JT, et al
(1992): Malonyl-CoA and long chain acyl-CoA esters as metabolic
coupling factors in nutrient-induced insulin secretion. J. Biol. Chem.
267, 5802 ± 5810.
Rosella G, Zaja JD, Baker L, Kaczmarczyk S, Antrikopoulos S, Adams TE
& Proietto J (1995): Impaired glucose tolerance and increased weight
gain in transgenic rats overexpressing a non-insulin responsive phos-
phoenolpyruvate carboxykinase gene. Mol. Endocrinol. 8, 1396 ± 1404.
Ruderman N & Herrera M (1968): Glucose regulation of hepatic gluco-
neogenesis. Am. J. Physiol. 214, 1346 ± 1351.
Ruderman N & Haudenschild C (1984): Diabetes as an atherogenic factor.
Prog. Cardiovasc. Dis. 26, 373 ± 412.
Ruderman NB, Toews CJ & Shafrir E (1969): Role of free fatty acids in
glucose homeostasis. Arch. Intern. Med. 123, 299 ± 313.
Russek M & Stevensen JAF (1972): Correlation between the effects of
several substances on food intake and on the hepatic concentration of
reducing sugars. Physiol. Behav. 8, 245 ± 249.
Russek M (1963): An hypothesis on the participation of hepatic gluco-
receptors in the control of food intake. Nature (London) 200, 176.
Russek M (1981): Current status of the hepatostatic theory of food intake.
Appetite 2, 137 ± 143 and 157 ± 162.
Schwarz J-M, Neese R, Shackleton CHL & Hellerstein MK (1993):
De novo lipogenesis (DNL) during fasting and oral fructose in lean and
obese hyperinsulinemic subjects. Diabetes 42, 39A (Suppl 1) (Abstr.).
Schwarz J-M, Neese RA, Turner S, Dare D & Hellerstein MK (1995):
Short-term alterations in carbohydrate energy intake in humans: striking
effects on hepatic glucose production, de novo lipogenesis, lipolysis and
whole-body fuel selection. J. Clin. Invest. 96, 2735 ± 2743.
Selmer J & Grunnet N (1976): Ethanol metabolism and lipid synthesis by
isolated liver cells from fed rats. Biochim. Biophys. Acta. 428, 123 ± 137.
Shimano H, Horton JD, Hammer R, Shimomura I, Brown M & Goldstein J
(1996): Overproduction of cholesterol and fatty acids causes massive
liver enlargement in transgenic mice overexpressing truncated SREBP-
1a. J. Clin. Invest. 98, 1575 ± 1584.
Shrago E, Glennon JA & Gordon ES (1971): Comparative aspects of
lipogenesis in mammalian tissues. Metabolism 20, 54 ± 62.
Shrago E, Spennetta T & Gordon E (1969): Fatty acid synthesis in human
adipose tissue. J. Biol. Chem. 244, 2761 ± 2766.
Sidossis LS & Wolfe RR (1996): Glucose and insulin-induced inhibition of
fatty acid oxidation: The glucose-fatty acid cycle reversed. Am. J.
Physiol. 270, E733 ± E738.
Siler S, Neese R & Hellerstein MK (1996): Effects of ethanol on lipolysis
and de novo lipogenesis in humans. FASEB J. 10, A799.
Sjostrom L (1973): Fatty acid synthesis de novo in adipose tissue from
obese subjects on a hypercaloric high-carbohydrate diet. Scand. J. Clin.
Lab. Invest. 32, 339 ± 349.
Storlein L, Pan D, Krisketos A, O'Connor J, Carson I, Covney G, Jenkins
A & Baur L (1996): Skeletal muscle membrane lipids and insulin
resistance. Lipids 31 Suppl 5, 261 ± 265.
Sul HS, Moustard N, Sakamoto K, Smas C & Gekakis N (1993):
Nutritional and hormonal regulation of genes encodengenzymes
involved in fat synthesis. In Nutrition and Gene Expression, eds. C
Berdanier, JL Hargrove. Baton Rouge: CRC, pp 207 ± 226.
Sullivan AC, Triscari J, Hamilton JG & Miller ON (1974): Effect of (-)-
hydroxycitrate upon the accumulation of lipid in the rat: II. Appetite.
Lipids 9, 129 ± 134.
Swinburn BA, Nyomba BL, Saad MF, Zurlo F, Raz I, Knowles WC,
Lillioja S, Bogardus C & Ravussin E (1991): Insulin resistance
associated with lower rates of weight gain in Pima Indians. J. Clin.
Invest. 88, 168 ± 173.
Tappy L, Paquot N, Tounian P, Schneiter P & Jequier E (1995): Assess-
ment of glucose metabolism in humans with the simultaneous use of
indirect calorimetry and tracer techniques. Clin. Physiol. 15, 1 ± 12.
Turner S, Jung H-R, Young S, Neese R & Hellerstein MK (1998):
Metabolic adaptations to the absence of dietary fat absorption and
chylomicrons in a transgenic-knockout mouse. FASEB J. 12, A514.
Valera A, Pujol A, Pelegrin M & Bosch F (1994): Transgenic mice
overexpressing phosphoenolpyruvate carboxykinase develop non-
insulin-dependent diabetes mellitus. Proc. Natl. Acad. Sci. USA 91,
9151 ± 9154.
Waterlow JC, Garlick PJ & Millward DJ, eds. (1978): Protein Turnover in
Mammalian Tissues and in the Whole Body. North Holland: Amster-
Weiss L, Hoffmann GE, Schreiber R, Andres H, Fuchs E, et al
(1986): Fatty acid biosynthesis in man, a pathway of minor import-
ance. Puri®cation, optimal assay conditions, and organ distribution
of fatty acid synthase. Biol. Chem. Hoppe-Seyler 367, 905 ±
Wing R, Blar H, Bononi P, Marcus MD, Watanabe R & Bergman RN
(1994): Caloric restriction per se is a signi®cant factor in improvements
in glycemic control and insulin sensitivity during weight loss in
NIDDM. Diabetes Care 17, 30 ± 36.
Young R & Sims EAH (1977): The woodchuck, Marmota monax, as a
laboratory animal. Lab. Animal Sci. 29, 771 ± 780.
Young S, Chan C, Pitas R, Burri B, Connolly A, Pappu A, Wong J,
Hamilton R & Farese R Jr (1995): A genetic model for absent
chylomicron formation: mice producing apolipoprotein B in the liver,
but not the intestine. J. Clin. Invest. 96, 2932 ± 2946.
De novo lipogenesis in humans