REPORTS

used a double-wall Couette apparatus with gap of 1

mm and measured the frequency dependence of the
storage modulus G!(") and loss modulus G#(") in a
small-amplitude oscillatory shear experiment.
22. Without preshearing, G!(") and G#(") indicated solid-like behavior similar to that reported for pure
smectics (7, 8) but with magnitudes of the moduli
three to four orders of magnitude smaller. This result
is consistent with the presence of a persistent macroscopically disordered texture of cholesteric layers.
23. Recall that a solid is characterized by G!(") 3

const $ 0 and G#(") 3 0 as " 3 0; this implies G!(")

% G#(") at sufficiently low frequencies. In a liquid, in
contrast, G!(")/G#(") 3 0 as " 3 0.
24. In addition to the measurements in the linear viscoelastic regime (u ! 0.05) described above, we
measured G!(",u) and G#(",u) at strain amplitudes u
up to 0.4 and found uG!(",u) to be practically uindependent for 0.2 ! u ! 0.4 and " ! 5 rad/s. The
low-frequency limit of uG!(",u) in this range of u is
the yield stress &y, which we measured to be 0.04
dyne. The strain value uy ! 0.15 at which uG!(",u)

Role of Nonexercise Activity

Thermogenesis in Resistance to
Fat Gain in Humans
James A. Levine, Norman L. Eberhardt, Michael D. Jensen*
Humans show considerable interindividual variation in susceptibility to weight
gain in response to overeating. The physiological basis of this variation was
investigated by measuring changes in energy storage and expenditure in 16
nonobese volunteers who were fed 1000 kilocalories per day in excess of
weight-maintenance requirements for 8 weeks. Two-thirds of the increases in
total daily energy expenditure was due to increased nonexercise activity thermogenesis (NEAT), which is associated with fidgeting, maintenance of posture,
and other physical activities of daily life. Changes in NEAT accounted for the
10-fold differences in fat storage that occurred and directly predicted resistance
to fat gain with overfeeding (correlation coefficient ' 0.77, probability ( 0.001).
These results suggest that as humans overeat, activation of NEAT dissipates
excess energy to preserve leanness and that failure to activate NEAT may result
in ready fat gain.
Weight gain occurs in healthy adults when
energy intake persistently exceeds energy expenditure. Some individuals appear to increase energy expenditure in response to
overeating without increasing volitional exercise and thus maintain a stable body weight.
This interindividual variation in weight gain
with overfeeding (1, 2) suggests that a thermogenic mechanism or mechanisms may be
activated to prevent weight gain or obesity.
When humans are overfed, more than 85%
of the stored excess energy is deposited as lipid
(3), primarily triglycerides. Lipid is ideally suited for long-term energy storage in mammals; it
is calorie-dense and hydrophobic, so that storage occurs without water accumulation. In the
presence of persistent, positive energy balance,
enormous quantities of triglyceride can be
stored through increases in adipocyte size and
number (4, 5). Even lean individuals store
enough fat to meet energy requirements for
J. A. Levine and M. D. Jensen, Department of Medicine,
Endocrine Research Unit, Mayo Clinic and Mayo Foundation, 200 First Street Southwest, Rochester, MN
55905, USA. N. L. Eberhardt, Departments of Medicine and of Biochemistry and Molecular Biology, Endocrine Research Unit, Mayo Clinic and Mayo Foundation, 200 First Street Southwest, Rochester, MN
55905, USA.
*To whom correspondence should be addressed. Email: [email protected]

212

more than 1 month, whereas some obese individuals have fat stores that would exceed energy requirements for a year (6, 7). However,
why some people appear to accumulate adipose
tissue more efficiently than others is unclear
(8).
The efficiency of energy storage is calculated by dividing the excess calories stored by the
excess calories consumed. Energy storage efficiency can never equal unity because heat transfer is not perfect. An energy efficiency of zero
would indicate that all excess energy consumed
is dissipated through increased energy expendi-

begins to saturate agrees well with &y /G!0 ! 0.2,

where G!0 ' 0.2 dyne/cm2 is the elastic shear modulus in the linear regime, thus providing an internal
consistency check of our data.
25. We thank T. Martin for participation in the early
stages of this project, V. Trappe for help with rheometry, and O. Lavrentovich for useful discussions.
This work was supported primarily through NSF grant
DMR95-07366, as well as the bourse Lavoisier du
Ministe
`re Francais des Affaires Etrange
`res.
11 September 1998; accepted 8 December 1998

ture. It has been argued that efficient energy

storage is beneficial because it allows longer
survival during famine. However, for many
Western populations, where food supply is
abundant and readily available, efficient energy
storage predisposes to obesity, the accumulation of excess body fat. Obesity affects more
than one-third of the U.S. population and is a
major public health concern because it is associated with diabetes, hypertension, hyperlipidemia, and cardiovascular disease (9).
Some humans appear to resist fat gain
with overeating, whereas others readily store
excess fat. These subjective observations
have been confirmed by a small number of
clinical studies that document a severalfold
interindividual variation in fat accumulation
with overfeeding (1, 2, 10). However, the
thermogenic adaptation that allows some individuals to resist weight gain despite overeating has not been identified.
To address this question, we designed a
study that allowed us to identify which component or components of energy expenditure
showed enough variability to account for the
variability in resistance to fat gain during overfeeding. Sixteen nonobese adults (12 males and
4 females, ranging in age from 25 to 36 years)
underwent measures of body composition and
energy expenditure before and after 8 weeks of
supervised overfeeding by 1000 kcal/day. Body
composition was measured with dual energy
x-ray absorptiometry (DXA) (11), and total daily energy expenditure was measured with doubly labeled water (10, 12). The latter procedure
required the administration of water containing

Table 1. Energy partitioning in 16 healthy human volunteers who were fed 1000 kcal/day (4.2 MJ) in
excess of weight maintenance requirements for 8 weeks. Additional data are available at www.
sciencemag.org/feature/data/982662.shl.
Variable (unit)

Mean

Range

Baseline weight (kg)

Overfed weight (kg)
Weight gain (kg)
Fat gain (kcal/day)*
Fat-free mass gain (kcal/day)*
Baseline dietary intake (kcal/day)
Baseline resting energy expenditure (kcal/day)
Overfed resting energy expenditure (kcal/day)
Baseline thermic effect of food (kcal/day)
Overfed thermic effect of food (kcal/day)
Baseline total energy expenditure (kcal/day)
Overfed total energy expenditure (kcal/day)

65.8
70.5
4.7
389
43
2824
1693
1772
218
354
2807
3361

53.391.7
58.8 93.1
1.4 7.2
58 687
1578
22653785
1470 1990
1460 2040
89 414
133 483
2216 3818
2508 4601

*Energy contents of tissues were calculated with published constants (3).

8 JANUARY 1999 VOL 283 SCIENCE www.sciencemag.org

REPORTS
isotopes of oxygen and hydrogen to the volunteers and measurement of the clearance of the
two isotopes from the body. The difference in
clearance of the two isotopes represents carbon
dioxide production (10, 12), which in turn reflects energy expenditure. These measurements
allowed us to observe how overeating affects
energy partitioning (Table 1). On average, 432
kcal/day of the excess energy ingested was
stored and 531 kcal/day was dissipated through
increased energy expenditure, thereby accounting for 97% of the additional 1000 kcal/day
(implying optimal compliance). Fat gain varied
10-fold among our volunteers, ranging from a
gain of only 0.36 kg to a gain of 4.23 kg, and
was inversely related to the increase in total
daily energy expenditure (r ' )0.86, P (
0.0001).
Total daily energy expenditure is composed of basal metabolic rate (BMR), postprandial thermogenesis, and physical activity
thermogenesis. BMR is the rate at which
energy is expended when an individual is
laying down at rest in the postabsorptive
state. We assessed BMR by using indirect
calorimetry to measure oxygen consumption
and carbon dioxide production (13). Changes
in BMR would be unlikely to account for the
10-fold variance in fat gain among our volunteers because previous investigators have
found only modest increases (*10%) with
overfeeding (10, 14). In our study, BMR
increased by an average of 5% in response to
overfeeding (Table 2), accounting for 8% of
the excess ingested energy. Thus, the interin-

dividual changes in BMR did not account for

the variability in fat gain (Fig. 1A).
Postprandial thermogenesis is the increase
in energy expenditure associated with the
digestion, absorption, and storage of food. It
may be the invariant energy cost of converting food to metabolic fuels (15, 16 ), or it may
be actively regulated in response to changing
food intake (17, 18). We measured postprandial thermogenesis using indirect calorimetry
and found that it increased by 14% with
overfeeding (Table 2). This increase was
more likely due to greater dietary intake (16 )
than to an adaptive response because the
thermic response to a meal of fixed energy
content (200 kcal, 0.8 MJ) was the same
before and after overfeeding (11 + 5 compared with 12 + 7 kcal per meal), consistent
with observations of other investigators (14).
Furthermore, interindividual differences in
postprandial thermogenesis did not correlate
with fat gain (Fig. 1B), suggesting that this
was not a significant factor in fat gain.
Physical activity thermogenesis can be subdivided into volitional exercise (sports and fitness-related activities) thermogenesis and what
we characterize as nonexercise activity thermogenesis (NEAT). NEAT is the thermogenesis
that accompanies physical activities other than
volitional exercise, such as the activities of
daily living, fidgeting, spontaneous muscle contraction, and maintaining posture when not recumbent. The possibility that NEAT might mediate resistance to fat gain intrigued us because
spontaneous physical activity (a component of

Fig. 1. The relation of the change in (A) basal metabolic rate, (B) postprandial thermogenesis, and
(C) activity thermogenesis with fat gain after overfeeding (2733). Exercise levels and the thermic
efficiency of exercise were unchanged with overfeeding, so that changes in activity thermogenesis
represent changes in NEAT.
Table 2. The fate of the excess 1000 kcal/day consumed by 16 volunteers during 8 weeks of overfeeding.
Data are expressed as kilocalories per day.
Variable

Mean

Standard deviation

Range

Fat mass gain*

Fat-free mass gain*
Change in resting energy expenditure
Change in thermic effect of food
Change in NEAT

389
43
79
137
328

188
22
126
83
256

58 687
1578
)100 360
28.2256
)98.3 692

*Energy contents of tissues were calculated with published constants (3).

NEAT) is a familial trait (19) that shows

marked interindividual differences in its contribution to daily energy expenditure (19, 20) and
is somewhat predictive of future weight gain
(21). Also, nonresting energy expenditure
(which includes NEAT) increases in adults subjected to a controlled 10% weight gain (22).
Finally, in previous overfeeding studies (3), it
has been possible to account for only *30% of
the calories that are wasted through increased
energy expenditure. If NEAT accounts for the
remaining 70%, then variable activation of
NEAT in response to overeating could explain
the wide variations in weight gain.
Measurement of overfeeding-induced changes in NEAT is formidable because of the
complexity of differentiating NEAT from volitional exercise thermogenesis in free-living
humans. We accomplished this differentiation by stringently maintaining volitional exercise at constant, low levels, and we confirmed compliance through questionnaires
and direct measures of physical activity. Although we appreciated that volitional exercise might change in response to overeating,
we viewed this as a behavioral rather than a
physiological adaptation and so elected to
eliminate it as a confounding variable. Because changes in exercise efficiency would
affect physical activity thermogenesis (23),
this variable was also measured. If the level
and efficiency of volitional exercise remained
constant over time, then changes in physical
activity thermogenesis (NEAT plus volitional
exercise) would represent changes in NEAT.
Hence, we assessed physical activity thermogenesis before and after overfeeding by measuring total daily energy expenditure using
doubly labeled water and subtracting from it
the sum of basal and postprandial energy
expenditure. These steps allowed us to assess
whether changes in NEAT mediate resistance
to fat gain with overfeeding.
NEAT proved to be the principal mediator
of resistance to fat gain with overfeeding. The
average increase in NEAT (336 kcal/day)
accounted for two-thirds of the increase in
daily energy expenditure (Table 2), and the
range of change in NEAT in our volunteers
was large ()98 to ,692 kcal/day). However,
most importantly, changes in NEAT directly
predicted resistance to fat gain with overfeeding (Fig. 1C), and this predictive value was
not influenced by starting weight (24).
Thus, activation of NEAT can explain the
variability in fat gain with overeating. As humans overeat, those with effective activation of
NEAT can dissipate the excess energy so that it
is not available for storage as fat, whereas those
with lesser degrees of NEAT activation will
likely have greater fat gain and be predisposed
to develop obesity. The maximum increase in
NEAT that we detected (692 kcal/day, volunteer 5) could be accounted for by an increase in
strolling-equivalent activity (25) by about 15

www.sciencemag.org SCIENCE VOL 283 8 JANUARY 1999

213

REPORTS
min/hour during waking hours. Of interest, the
four lowest values for change in NEAT corresponded to the four female volunteers, although
the relation between change in NEAT and fat
gain was the same in males and females. A
larger study will be needed to determine the
significance of the preliminary gender differences noted here. Another limitation of our
study is the small errors inherent in measuring
energy expenditure and body composition in
physiological studies. Because these errors are
cumulative, they would be expected to weaken
the association between the change in NEAT
and the change in body fat. Thus, it is possible
that we have underestimated the contribution of
NEAT activation to the resistance to fat gain
with overfeeding.
Finally, our results suggest that efforts to
enhance NEAT activation, perhaps through
behavioral cues, may be a fruitful approach to
the prevention of obesity.

28.

References and Notes

1. E. A. Sims et al., Recent. Prog. Horm. Res. 29, 457

(1973).
2. C. Bouchard et al., N. Engl. J. Med. 322, 1477 (1990).
3. O. Deriaz, A. Tremblay, C. Bouchard, Obes. Res. 1, 179
(1993).
4. J. B. Prins and S. ORahilly, Clin. Sci. 92, 3 (1997).
5. J. A. Levine, M. D. Jensen, N. L. Eberhardt, T. OBrien,
J. Clin. Invest. 101, 1557 (1998).
6. A. E. Black, W. A. Coward, T. J. Cole, A. M. Prentice,
Eur. J. Clin. Nutr. 50, 72 (1996).
7. G. F. Cahill, N. Engl. J. Med. 282, 668 (1970).
8. E. A. Sims and E. Danforth, J. Clin. Invest. 79, 1019
(1987).
9. W. P. James, Int. J. Obes. Relat. Metab. Disord. 16,
S23 (1992).
10. E. O. Diaz, A. M. Prentice, G. R. Goldberg, P. R.
Murgatroyd, W. A. Coward, Am. J. Clin. Nutr. 56, 641
(1992).
11. B. M. Prior et al., J. Appl. Physiol. 83, 623 (1997).
12. W. A. Coward, S. B. Roberts, T. J. Cole, Eur. J. Clin.
Nutr. 42, 207 (1988).
13. E. Jequier and J. P. Felber, Baillieres Clin. Endocrinol.
Metab. 1, 911 (1987).
14. A. Tremblay, J. P. Despres, G. Theriault, G. Fournier, C.
Bouchard, Am. J. Clin. Nutr. 56, 857 (1992).
15. J. O. Hill, S. B. Heymsfield, C. D. McMannus, M.
DiGirolamo, Metabolism 33, 743 (1984).
16. D. A. Alessio et al., J. Clin. Invest. 81, 1781 (1988).
17. K. R. Segal et al., ibid. 89, 824 (1992).
18. E. Ravussin, B. Burnand, Y. Schutz, E. Jequier, Am. J.
Clin. Nutr. 41, 753 (1985).
19. E. Ravussin, S. Lillioja, T. E. Anderson, L. Christin, C.
Bogardus, J. Clin. Invest. 78, 1568 (1986).
20. S. Toubro, N. J. Christensen, A. Astrup, Int. J. Obes.
Relat. Metab. Disord. 19, 544 (1995).
21. F. Zurlo et al., Am. J. Physiol. 263, E296 (1992).
22. R. L. Leibel, M. Rosenbaum, J. Hirsch, N. Engl. J. Med.
332, 621 (1995).
23. J. Kang et al., Med. Sci. Sports Exercise 29, 377
(1997).
24. If total daily energy expenditure measured with doubly labeled water is assumed to equal weight maintenance requirements (rather than the measures of
weight maintenance dietary intake), the relation between the increase in NEAT and the efficiency of
energy storage (excess kilocalories stored/number of
excess kilocalories provided) is almost identical to the
relation we report in Fig. 1C (r ' )0.80, P ( 0.001,
compared with r ' )0.77, P ( 0.001).
25. N. G. Norgan and J. V. Durnin, Am. J. Clin. Nutr. 33,
978 (1980).
26. W. A. Coward, Proc. Nutr. Soc. 47, 209 (1988).
27. The 16 (12 males and 4 females) healthy volunteers
were 25 to 36 years old. Volunteers were excluded if
they used any medication at the time of the study or

214

29.

30.

31.

within 6 months of the study, exercised more than

twice each week, smoked, used alcohol, were pregnant, had any acute or chronic illness, or reported
unstable body weight.
Volunteers were studied as outpatients for 10 weeks.
Meals were prepared in the metabolic kitchen at the
Mayo Clinic General Clinical Research Center (GCRC).
All foods were weighed to within 1 g. For the first 2
weeks, volunteers were fed so as to establish the dietary
intake necessary to maintain steady-state body weight.
For the remaining 8 weeks, each volunteer received
1000 kcal in addition to weight maintenance requirements. The diet composition remained constant
throughout the study at 40% carbohydrate, 40% fat,
and 20% protein. The volunteers body weight was
measured each morning under standardized conditions
(with an empty bladder, without shoes, and wearing
consistent, light clothing); these measures were made
by GCRC personnel. Volunteers were instructed not to
adopt new exercise practices and were questioned daily
regarding activities. In addition, volunteers family and
friends underwent structured interviews before and after feeding to determine compliance with exercise restrictions. During weeks 2 and 10, volunteers wore
accelerometers (with disabled liquid crystal displays)
(Caltrac; Muscle Dynamics, Torrance, CA) to measure
the extent of free-living exercise-related activity. To
ensure compliance with the feeding regimen, volunteers
were instructed to eat all foods provided, and almost all
meals were consumed under supervision at the GCRC.
Plates were inspected for solid or liquid remainders.
When food items were eaten outside of the GCRC,
preweighed food items were provided by the investigators, and empty food containers were inspected. On
occasion, volunteers home garbage was checked. Family members, friends, and work colleagues of the volunteers were identified and contacted on several occasions throughout the study to ensure that all food was
consumed and that exercise was not initiated. Informed
consent was obtained after the nature and possible
consequences of the study were explained.
Each volunteer was weighed daily with the same calibrated scale. Body fat and mineral mass were measured
in duplicate with DXA after baseline feeding (end of
week 2) and after completion of overfeeding (end of
week 10). To ensure that our measures of body composition were reproducible and precise, (i) we used the
same DXA scanner throughout the study, (ii) we calibrated the DXA scanner before each measurement with
tissue phantoms, and (iii) we calibrated the DXA scanner against tissue blocks of known composition weekly.
A human adipose tissue block with a lipid content of
2891 g by chemical analysis was found to be 2949 g by
DXA scans. Comparison of fat-free mass obtained with
the DXA and isotope dilution revealed a strong correlation (r ' 0.97, P ( 0.0001). Finally, when a 600-g
block of adipose tissue was placed on a volunteer with
22.8 kg of body fat as assessed by DXA, 577 g of this
block was detected. Fat-free mass was calculated from
the difference between body weight and fat mass. The
test-retest difference for duplicate measurements was
(2%.
BMR was measured on two consecutive mornings at
0630 in volunteers who had slept uninterrupted the
previous nights in the GCRC. Volunteers were not
moved before measurements and had not eaten since
2100 the night before. For each measurement, the
calorimeter (Deltatrac; SensorMedics, Yorba Linda, CA)
was calibrated with gases of known composition. Volunteers were awake, semirecumbent (10 head bed tilt),
lightly clothed, and in thermal comfort (68 to 74F) in
a dimly lit, quiet room. Measurements were performed
for 30 min during which time volunteers were not
allowed to talk or move. The test-retest difference for
duplicate measurements was (3%.
Postprandial thermogenesis was measured on two
consecutive days at the end of weeks 2 and 10. On
the first study day, volunteers were given a meal that
provided one-third of their daily intake (40% carbohydrate, 40% fat, and 20% protein). Energy expenditure was measured with the indirect calorimeter for
15 of every 30 min (to prevent agitation) until values
within 4 kcal/hour of resting energy expenditure
were recorded for two consecutive measurements.
On the second day, volunteers were provided with a

200-kcal meal (40% carbohydrate, 40% fat, and 20%

protein), and the same procedures were followed.
Areas under the curves for time (x axis) and energy
expenditure ( y axis) were used to determine postprandial thermogenesis. The mean duration of measurement was 414 + (SD) 39 min. Daily postprandial
thermogenesis was calculated by tripling the postprandial thermogenesis obtained after the meal providing one-third of daily intake.
32. Total energy expenditure was measured in weeks 2 and
10 with doubly labeled water (12, 26). Baseline urine
samples were collected, and after timed administration
of the isotopes, urine samples were collected at 0700,
1200, and 1800 each day for 7 days. The slope-intercept
equations described by Coward et al. (12) were used to
derive values for total energy expenditure. Propagation
or error analysis was performed (10) on each measurement, and the calculated compounded errors (measurement plus biological noise) were 3 + 1% for baseline
and 4 + 3% after overfeeding. Measures of baseline
total energy expenditure derived with doubly labeled
water were in excellent agreement with the measures
of baseline weight-maintenance dietary intake (r '
0.89, P ( 0.001, with an intercept not different from 0
and a slope not different than 1).
33. Changes in NEAT were measured by calculating activity-related thermogenesis before and after overfeeding.
Activity-related thermogenesis was determined by
measuring total energy expenditure, with doubly labeled water, and subtracting from it the sum of basal
energy expenditure and postprandial energy expenditure. Subtraction of the value for activity-related thermogenesis before overfeeding from the value obtained
after overfeeding represented the change in NEAT if
two conditions were met: (i) the total amount of volitional exercise was unchanged and (ii) the thermic
efficiency of exercise was unchanged. We determined
that the amount of exercise did not change with overfeeding on the basis of accelerometer readings [2905 +
(SD) 514 accelerometer units (AU)/day before overfeeding compared with 2963 + 540 AU/day after overfeeding], daily interviews with the volunteers regarding
their exercise level, and structured interviews with volunteers and their relatives and friends before and after
overfeeding. The thermic efficiency of exercise was
assessed with two different measures of exercise-related energy expenditure at the end of weeks 2 and 10.
The first measure was during treadmill walking at 3
mph (4.8 km/hour) for 10 min. Throughout this period,
inspired and expired gases were sampled for volume O2
and CO2 content with an integrated treadmill and gas
sampling mass spectrometer system. Mean VO2 for
minutes 3 to 9 was calculated. The volunteers also
bicycled on a stationary bicycle at 100 W for 20 min.
Between minutes 3 to 5 and 13 to 15, exhaled air was
collected with a three-way mouth piece and leak-proof
bag. The volume of expired air for the 2-min period was
measured with a Tissot spirometer, and the O2 and CO2
concentrations were measured with a calibrated mass
spectrometer. Gas volumes were corrected for standard
temperature and pressure and humidity. Average oxygen consumption (VO2) for the two bag collections was
calculated. The thermic responses to cycling and walking did not change with overfeeding; before overfeeding, VO2 during cycling at 100 W was 1693 + 39
ml/min, and after overfeeding it was 1772 + 43 ml/
min; before overfeeding, VO2 during walking at 3 mph
was 1028 + 59 ml/min, and after overfeeding it was
1061 + 38 ml/min. Thus, because volitional exercise
and the thermic efficiency of exercise were unchanged
with overfeeding, any change in activity-related thermogenesis after overfeeding represented the change in
NEAT. Finally, to ensure that energy wastage did not
occur through malabsorption, 3-day stool fat was measured before and after overfeeding. There was no significant increase in stool fat after overfeeding (25 + 13
kcal/day compared with 38 + 15 kcal/day).
34. We thank the volunteers, dietitians, food technicians,
and nursing staff at the GCRC and A. Wright and
W. A. Coward for assistance with doubly labeled water
calculations. Supported by NIH grants DK45343,
DK50456, and M01 RR00585 and the Mayo Foundation.
21 May 1998; accepted 13 November 1998

8 JANUARY 1999 VOL 283 SCIENCE www.sciencemag.org