Dynamics of fat cell turnover in humans
Kirsty L. Spalding1, Erik Arner1, Pa ˚l O. Westermark2, Samuel Bernard3, Bruce A. Buchholz4, Olaf Bergmann1,
Lennart Blomqvist5, Johan Hoffstedt5, Erik Na ¨slund6, Tom Britton7, Hernan Concha5, Moustapha Hassan5,
Mikael Ryde ´n5, Jonas Frise ´n1& Peter Arner5
constitutes a public health problem by enhancing the risk for
cardiovascular disease and metabolic disorders such as type 2 dia-
betes1,2. Owing to the increase inobesity, life expectancy maystart
to decrease in developed countries for the first time in recent
history3. The factors determining fat mass in adult humans are
not fully understood, but increased lipid storage in already
developed fatcells(adipocytes) isthoughttobemostimportant4,5.
Here we show that adipocyte number is a major determinant for
the fat mass in adults. However, the number of fat cells stays
constant in adulthood in lean and obese individuals, even after
marked weight loss, indicating that the number of adipocytes is
set during childhood and adolescence. To establish the dynamics
within the stable population of adipocytes in adults, we have
measured adipocyte turnover by analysing the integration of
14C derived from nuclear bomb tests in genomic DNA6.
Approximately 10% of fat cells are renewed annually at all adult
ages and levels of body mass index. Neither adipocyte death nor
generation rate is altered in early onset obesity, suggesting a tight
regulation of fat cell number in this condition during adulthood.
The high turnover of adipocytes establishes a new therapeutic
target for pharmacological intervention in obesity.
The fat mass can expand by increasing the average fat cell volume
and/or the number of adipocytes. Increased fat storage in fully dif-
ferentiated adipocytes, resulting in enlarged fat cells, is well docu-
fat depots increase in adults4,5. To analyse the contribution of the fat
cell volume in adipocytes to the size of the fat mass, we first analysed
the relationship between fat cell volume and total body fat mass
(directly measured with bioimpedance or estimated from body mass
index (BMI), sex and age in a large cohort of adults). As expected,
about 80% of all fat, and in visceral fat (Fig. 1d), which has a strong
between fat cell volume and fat mass markedly differed from a linear
relationship (likelihood ratio test P,0.001, and Akaike information
criterion, described in Supplementary Information 1) in both sub-
cutaneous and visceral adipose regions and both sexes, indicating
that fat mass is determined by both adipocyte number and size. In
only important determinant of fat mass.
adipose tissue during childhood7, but it is unknown whether the
number of adipocytes changes during adulthood. We assessed the
total adipocyte number in 687 adult individuals and combined this
data with previously reported results for children and adolescents8.
Although the total adipocyte number increased in childhood and
hood in both lean and obese individuals (adults over 20yr, grouped
in 5-yr bins; ANOVA, lean P50.68, obese P50.21; Fig. 2a and
Supplementary Information 3). Thus, the difference in adipocyte
number between lean and obese individuals is established during
childhood7,8and the total number of adipocytes for each weight
category stays constant during adulthood (Fig. 2b). The small vari-
ation in adipocyte number for each BMI category demonstrates that
this is a stable cell population during adulthood.
To analyse whether alterations in adipocyte number may contri-
bute to changed fat mass under extreme conditions, we next asked
whether fat cell number is reduced during major weight loss (mean
body weight loss, 18611%, mean6s.d.) by radical reduction in
calorie intake by bariatric surgery (reduction of the stomach with
the purpose of facilitating weight loss). The surgical treatment
this failed to reduce adipocyte cell number two years post surgery
(Fig. 2b, c and Supplementary Information 4), in line with previous
a complementary longitudinal study13. Ref. 13 found that significant
weight gain (15–25%) over several months in non-obese adult men
resulted in a significant increase in body fat, which was accompanied
by an increase in adipocyte volume, but no change in adipocyte
number. Similar to our findings, subsequent weight loss back to
baseline resulted in a decrease of adipocyte volume, but, again, no
change in adipocyte number. Although we cannot rule out that a
more prolonged period of weight gain in adulthood could result in
an increase in adipocyte number, these results and ours indicate that
mass in adulthood can mainly be attributed to changes in fat cell
volume. This may indicate that the number of adipocytes is set by
early adulthood with no subsequent cell turnover. Alternatively, the
generation of adipocytes may be balanced by adipocyte death, with
the total number being tightly regulated and constant.
We next set out to establish whether adipocytes are replaced dur-
ing adulthood, and, if so, at what rate. Adipocytes can be generated
from adult human mesenchymal stem cells and pre-adipocytes
whether adipocytes are generated in vivo14. Cell turnover has been
difficult to study in humans. Methods used in experimental animals,
such as the incorporation of labelled nucleotides, cannot readily be
adapted for use in humans owing to potential toxicity. The detection
of cells expressing molecular markers of proliferation can give
1Department of Cell and Molecular Biology, Karolinska Institute, SE-171 77 Stockholm, Sweden.2Institute for Theoretical Biology (ITB), Humboldt University Berlin and Charite ´,
Accelerator Mass Spectrometry, Lawrence Livermore National Laboratory, 7000 East Avenue, L-397, Livermore, California 94551, USA.5Department of Medicine, Karolinska
University Hospital, SE-141 86 Stockholm, Sweden.6Division of Surgery, Department of Clinical Science, Danderyds Hospital, Karolinska Institutet, SE-182 88 Stockholm, Sweden.
7Department of Mathematics, Stockholm University, 106 91 Stockholm, Sweden.
Vol 453|5 June 2008|doi:10.1038/nature06902
insights about mitotic activity, but fail to provide information
regarding the fate of the progeny of the dividing cells. This is a
limitation when studying postmitotic cell types, which do not divide
or express mitotic markers themselves (for example, neurons or
adipocytes) but may be replenished from proliferating stem or pro-
genitor cells, such as preadipocytes.
and allowsthe analysis of cellturnover inhumans6,18. Levels of14C in
the atmosphere were relatively stable until the Cold War, when
above-ground nuclear bomb tests (1955–1963) caused a notable
at a limited number of locations, increased14C levels in the atmo-
sphere rapidly equalized around the globe. Since theTest-Ban Treaty
in 1963, the14C levels have dropped exponentially, not because of
sphere21. Atmospheric14C reacts with oxygen to form CO2, which is
incorporated into plants by photosynthesis. By eating plants, and
Because DNA is stable after a cell has gone through its last cell divi-
sion, the14C level in DNA serves as a date mark for when a cell was
To address whether adipocytes are generated from newborn cells
in adulthood, we isolated fat cells ($98% purity, Supplementary
Information 4) from adipose tissue collected during liposuction or
abdominal wall reconstruction from 35 adult lean or obese indivi-
duals. The pure isolation of adipocytes is important because non-
adipose cells are present in adipose tissue and these cell types may
full discussion). Genomic DNA was extracted from the purified
spectrometry and related to atmospheric14C data (Fig. 3c, d and
Supplementary Information 4). We first analysed individuals born
14C levels were measured by accelerator mass
Fat cell volume
Fat cell number (×1010)
Figure 2 | Adipocyte number remains stable in adulthood, although
significant weight loss can result in a decrease in adipocyte volume. Total
adipocyte number from 595 (n lean5253; n obese5342) adult individuals
(squares) was combined with previous results for children and adolescents8
(circles; n lean5178; n obese5120). a, The adipocyte number increases in
childhood and adolescence, with the number levelling off and remaining
constant in adulthood in both lean (blue) and obese (pink) individuals.
cell volume (b), however fails to reduce adipocyte cell number (c), 1–2yr
post surgery (n520). All error bars represent s.e.m.; asterisk, P,0.0001.
Fat cell volume (picolitres)
0 2040 6080100120
Body fat mass (kg)
02040 6080 100120
Figure 1 | Fat mass is determined by both adipocyte number and size.
a–d, The relationship between fat mass and fat cell volume was curvilinear
female584, n male551). This demonstrates that both adipocyte number
and adipocyte size are determinates of body fat mass. In a, b and d, body fat
mass was estimated from BMI using a previously described formula
c, fat mass was determined using bioimpedance. Fat cell volume is given in
picolitres, where 10212litres51029cm3.
NATURE|Vol 453|5 June 2008
well before the period of nuclear bomb tests. This provides a high
of the nuclear bomb tests (1955), because14C levels above those
present before the Cold War can be detected even if only a small
individuals born before 1955 (n510), the14C levels were substan-
tially higher than the atmospheric levels before the nuclear bomb
tests, indicating that generation of adipocytes had taken place after
1955 (Fig. 3c, see Supplementary Information 2 for all14C measure-
ments and associated data). The individuals were 0–22years old at
the onset of nuclear bomb tests, establishing that adipocytes are
generated during adolescence and in early adulthood. New adipo-
cytes may also be formed by differentiation of existing post-mitotic
pre-adipocytes; hence, DNA integration of
bound to the generation of adipocytes.
Analysis of individuals born before the onset of the nuclear bomb
tests provides a high sensitivity to detect cell turnover, but alone does
not allow the establishment of the turnover rate becausea certain14C
14C curve. However, the integration of data from individuals born
fore also analysed14C levels in adipocyte genomic DNA from indivi-
duals born after the period of nuclear bomb tests (n525). In all of
porary time points (Fig. 3d and Supplementary Information 2), pro-
viding a first indication that there is continuous and substantial
turnover of adipocytes in adult humans.
14C provides a lower
Wenext calculated thedynamics offatcell turnover using a simple
birth and death model (detailed in Supplementary Information 3).
The model’s assumptions allow the calculation of kinetic rates for
individual subjects. The death rate of adipocytes is approximately
8.466.2% per yr (median6average deviation) in the total fat pool
of the body. The distribution of death rates is skewed towards lower
P,0.05);therefore,themedian6averagedeviation is more inform-
ative than the mean6s.d.25. To test the reliability of the death-rate
death rates do not differ from the median (sign test, P.0.3; see
Supplementary Information 3 for description of the scenarios). We
divided the data set into lean (BMI,25kg per m2) and obese
(BMI$30kg per m2, all of which had early onset obesity, see
on adipocyte death rate. No significant difference in adipocyte death
rate was seenacrossthedifferentBMIs,with obeseindividualshaving
ation) per yr, versus 8.265.3% (median6average deviation) per yr
average cell number in subjects aged 20–70yr using data presented in
Fig. 2b (n5650 and P50.19 by linear regression analysis), arguing
that the adipocyte death rate per yr must be matched with a similar
weight categories. We calculate a median turnover rate of 8.466.2%
(median6average deviation) per yr, with half of the adipocytes
replaced every 8.3yr.
Fraction modern (F)
Fraction modern (F)
Fraction modern (F)
1925 1935 1945 1955 1965 1975 1985 1995 2005
1925 1935 1945 1955 1965 1975 1985 1995 2005
1925 1935 1945 1955 1965 1975 1985 1995 2005
1500BC1000BC 500BC0 5001000 15002000
Figure 3 | Turnoverofadipocytesinadulthood. a,b,Thelevelsof14Cinthe
atmosphere have been relatively stable over long time periods, with the
tests21. The boxed region in a is shown in more detail in b.14C levels from
modern samples are by convention given in relation to a universal standard
and corrected for radioactive decay, giving the D14C value30. c, d, Adipocyte
were analysed by determining the14C concentration in adipocyte genomic
DNA using accelerator mass spectrometry. The measured14C value is
related to the recorded atmospheric levels to establish at what time point
they corresponded. The year is plotted on the x axis, giving the birth date of
the cell population. Three representative individuals born at different times
before the onset of the bomb tests reveals the generation of adipocytes after
birth (c). Analysis of the oldest individuals established that adipocytes are
after the period of nuclear bomb tests showed continuous and substantial
indicated by a vertical line in each graph and the BMI is shown numerically
(c, d). Error bars for the accelerator mass spectrometry readings are too
small to be visualized in this graph. Each dot represents one individual.
NATURE|Vol 453|5 June 2008
Using the death-rate estimates and the fat cell numbers calculated
for individual subjects, absolute fat cell production was calculated.
Obese individuals were found to have a significantly greater number
cells per yr versus (0.360.2)31010cells per yr (median6average
deviation; P,0.01, ANOVA; Fig. 4b). Loss of fat cells is therefore
in obese subjects compared with lean subjects. The fact that the total
number of new adipocytes added each year is greater in obese com-
pared with lean individuals, yet the proportion of newborn adipo-
cytes added each year (the turnover) is the same for both groups,
argues that the difference in cell number between the lean and obese
adults occurs before adulthood. In support of this, we found no
significant difference in the average age of adipocytes in lean
9.963.5yr (mean6s.d.) versus obese 9.764.0yr (mean6s.d.)
individuals (Fig. 4c). No significant correlation between the age of
subjects and cell death or between the age of patients and adipocyte
generation was found (Supplementary Information 3), suggesting a
constant turnover rate throughout adult life.
If the number of adipocytes is set to a higher level in obese people
before adulthood, this could be because cell-number expansion
begins earlier (age of onset), because expansion is faster (growth
relative to the initial cell number (IC) at age of onset), or because
expansion ends later (age at 90% of adult cell number). We used
combined adipocyte number data (Fig. 2a) to see whether one or
more of these factors determine adipocyte number. Using our birth
anddeath model, wedetermined that age at onset ofadipocyte num-
ber expansion is significantly earlier in obese (2.160.9yr) than in
lean(5.760.8yr)subjects;therelativeincrease inadipocyte number
is higher in obese (2.460.6ICyr–1) than in lean (1.360.3ICyr–1)
(16.561.3yr) than in lean (18.560.7yr) subjects (all values are
predicted values695% confidence interval, Supplementary Infor-
and is not caused by a prolonged expansion period in adulthood.
We find that the number of adipocytes for lean and obese indivi-
duals is set during childhood and adolescence, and that adipocyte
numbers for these categories are subject to little variation during
adulthood. Even after significant weight loss in adulthood and
reduced adipocyte volume, the adipocyte number remains the same.
Although we show that the adipocyte number is static in adults, we
alsodemonstrate that thereisremarkable turnover withinthispopu-
influenced by the energy balance. Studies of previously obese indivi-
duals after weight loss show that their adipose tissue hypercellularity
andto lowerenergy expenditure26.These factors promote lipidaccu-
loss. Thus, a tight regulation of adipocyte number, together with
mechanisms maintaining their energy balance, may contribute to
why obese individuals have difficulties maintaining weight loss.
onset of obesity. We cannot rule out that those who gradually gain
adipocyte size until a threshold is reached and thereafter recruit new
fat cells from committed precursor cells or mesenchymal stem cells.
Most obese adults have been obese since childhood, with less than
10% of children with normal weight going on to develop adult
obesity27. By contrast, over three-quarters of obese children go on
to become obese adults27. Thus, understanding the dynamics of
adipocyte turnover in adults who have been obese since childhood
is of great importance, especially given the current trend for an
increase in childhood obesity.
The size of organs can be regulated by different mechanisms, and
the number of cells in some tissues is controlled by a systemic feed-
back mechanism28. This is best understood for skeletal muscle, in
which growth and differentiation factor 8 (GDF8), also known as
myostatin, is secreted from myocytes and negatively regulates the
generation of new muscle cells and thereby sets the number of cells29.
Loss-of-function mutations in GDF8 result in a large increase in the
number (and size) of myocytes in animals and humans29. The steady
production of adipocytes in adults results in a stable size of the con-
stantly turning over adipocyte population. Feedback mechanisms
that control adipocyte turnover will be important to identify at a
molecular level because this may offer a novel target for pharmaco-
logical therapy when obesity is established and for other types of
intervention during childhood and adolescence when the final num-
ber of fat cells in the body is being set.
Subjects. The relationship between subcutaneous or visceral fat cell volume,
BMI and fat mass was studied in two separate cohorts, and fat cell turnover
was studied in a third cohort, all of which are described in Supplementary
Isolated fat cells. Fat cells were isolated from the adipose tissue as described in
Supplementary Information 4. Details on how to measure weight, volume and
given in Supplementary Information 4.
14C analysis. Genomic DNA was prepared from isolated fat cells, and was puri-
fied and subjected to accelerator mass spectrometry analyses, as described in
Supplementary Information 4 and tabled in Supplementary Information 2.
Dataanalysis.The calculationsof relationshipbetweenfat cell volumeand BMI
or fat mass are described in detail in Supplementary Information 1. Thecalcula-
tions of fat cell death and generation are described in detail in Supplementary
Received 30 November 2007; accepted 7 March 2008.
Published online 4 May 2008.
1.Van Gaal, L. F., Mertens, I. L. & De Block, C. E. Mechanisms linking obesity with
cardiovascular disease. Nature 444, 875–880 (2006).
Kahn, S. E., Hull, R. L. & Utzschneider, K. M. Mechanisms linking obesity to insulin
resistance and type 2 diabetes. Nature 444, 840–846 (2006).
Olshansky, S. J. et al. A potential decline in life expectancy in the United States in
the 21st century. N. Engl. J. Med. 352, 1138–1145 (2005).
Death rate (per yr)
(×1010 cells per yr)
Adipocyte age (yr)
Figure 4 | Effect of obesity on adipocyte generation and death. a, No
significant difference in adipocyte death rate per year was seen across the
different BMIs. b, Obese individuals had a significantly greater number of
the average age of adipocytes in lean versus obese individuals was found. In
and third quartiles; in c, data are shown as mean6s.e.m. Asterisk, P,0.01
for lean (n513) versus obese (n514) individuals, Kruskal–Wallis test.
NATURE|Vol 453|5 June 2008
4. Bjorntorp, P. Effects of age, sex, and clinical conditions on adipose tissue
cellularity in man. Metabolism 23, 1091–1102 (1974).
Hirsch, J. & Batchelor, B. Adipose tissue cellularity in human obesity. Clin.
Endocrinol. Metab. 5, 299–311 (1976).
birth dating of cells in humans. Cell 122, 133–143 (2005).
Prins, J. B. & O’Rahilly, S. Regulation of adipose cell number in man. Clin. Sci.
(Lond.) 92, 3–11 (1997).
of adipose tissue in children and adolescents. Cross-sectional and longitudinal
studies of adipose cell number and size. J. Clin. Invest. 63, 239–246 (1979).
biochemical and radioautographic techniques. J. Lipid Res. 25, 336–347 (1984).
10. Kral, J. et al. Body composition and adipose tissue cellularity before and after
jejuno-ileostomy in severely obese subjects. Eur. J. Clin. Inv. 7, 413–419 (1977).
tissue cellularity in obese women. Am. J. Clin. Nut. 28, 445–452 (1975).
12. Ha ¨ger, A. et al. Adipose tissue cellularity in obese school girls before and after
dietary intervention. Am. J. Clin. Nut. 31, 68–75 (1978).
13. Sims, E. A. et al. Experimental obesity in man. Trans. Assoc. Am. Physicians 81,
14. Rodriguez, A. M., Elabd, C., Amri, E. Z., Ailhaud, G. & Dani, C. The human adipose
tissue is a source of multipotent stem cells. Biochimie 87, 125–128 (2005).
15. Petruschke, T. & Hauner, H. Tumor necrosis factor-a prevents the differentiation
of human adipocyte precursor cells and causes delipidation of newly developed
fat cells. J. Clin. Endocrinol. Metab. 76, 742–747 (1993).
16. Prins, J. B., Walker, N. I., Winterford, C. M. & Cameron, D. P. Apoptosis of human
adipocytes in vitro. Biochem. Biophys. Res. Commun. 201, 500–507 (1994).
17. Cinti, S. et al. Adipocyte death defines macrophage localization and function in
adipose tissue of obese mice and humans. J. Lipid Res. 46, 2347–2355 (2005).
18. Bhardwaj, R. D. et al. Neocortical neurogenesis in humans is restricted to
development. Proc. Natl Acad. Sci. USA 103, 12564–12568 (2006).
19. De Vries, H. Atomic bomb effect: variation of radiocarbon in plants, shells, and
snails in the past 4 years. Science 128, 250–251 (1958).
hemisphere (1959–2003). Radiocarbon 46, 1261–1272 (2004).
22. Spalding, K. L., Buchholz, B. A., Bergman, L. E., Druid, H. & Frisen, J. Forensics: age
written in teeth by nuclear tests. Nature 437, 333–334 (2005).
for human tissue from atmospheric radiocarbon. Science 146, 1170–1172 (1964).
24. Harkness, D. D. Further investigations of the transfer of bomb14C to man. Nature
240, 302–303 (1972).
25. Altman, D. G. Practical Statistics for Medical Research pp 164 (Chapman & Hall,
CRC, London, 1991).
26. Lo ¨fgren, P. et al. Long-term prospective and controlled studies demonstrate
adipose tissue hypercellularity and relative leptin deficiency in the post-obese
state. J. Clin. Endocrinol. Metab. 90, 6207–6213 (2005).
27. Freedman D. S. . et al. Relationship of childhood overweight to coronary heart
disease. Risk factors in adulthood: The Bogalusa Heart Study. Pediatrics 108,
26, 173–175 (1996).
29. Joulia-Ekaza, D. & Cabello, G. Myostatin regulation of muscle development:
molecular basis, natural mutations, physiopathological aspects. Exp. Cell Res. 312,
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank M. Stahlberg and T. Bergman for help with
high-performance liquid chromatography (HPLC), D. Kurdyla, P. Zermeno and
A. Williams for producing graphite, and S. Zdunek for comments on the statistics
and modelling. This study was supported by grants from Knut och Alice
Wallenbergs Stiftelse, the Human Frontiers Science Program, the Swedish
Research Council, the Swedish Cancer Society, the Swedish Heart and Lung
foundation, the Novo Nordic Foundation, the Swedish Diabetes Foundation, the
Foundation forStrategicResearch, theKarolinskaInstitute, theTobias Foundation,
AFA Life Insurance Health Foundation and NIH/NCRR (RR13461). This work was
of California, Lawrence Livermore National Laboratory under contract
Author Contributions K.L.S., P.A. and J.F. designed the study and wrote the
statistics. K.L.S. and B.A.B. performed sample preparation and14C accelerator
mass spectrometry measurements. L.B., J.H. and E.N. collected clinical material.
H.C., M.H. and M.R. performed studies on fat cell purity.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. Correspondence and requests for materials should be
addressed to K.L.S. (email@example.com), J.F. (firstname.lastname@example.org) or P.A.
NATURE|Vol 453|5 June 2008