The turnover of adipose tissue components (lipids and cells) and the pathways of adipose lipid deposition have been difficult to measure in humans. We apply here a 2H2O long-term labeling technique for concurrent measurement of adipose-triglyceride (TG) turnover, cell (DNA) proliferation, and de novo lipogenesis (DNL). Healthy subjects drank 2H2O (70 ml/day) for 5-9 wk. Subcutaneous adipose tissue aspirates were taken (gluteal, thigh, and flank depots). Deuterium incorporation into TG glycerol (representing all-source TG synthesis), TG palmitate (representing DNL, by mass isotopomer distribution analysis), and DNA (representing cell proliferation) was measured by gas chromatography-mass spectrometry. Subjects tolerated the protocol well, and body 2H2O enrichments were stable. Mean TG-glycerol fractional synthesis was 0.12 (i.e., 12%) with a range of 0.03-0.32 after 5 wk and 0.20 (range 0.08-0.49) after 9 wk (TG half-life 200-270 days). Label decay measurements 5-8 mo after discontinuing 2H2O gave similar turnover estimates. Net lipolysis (TG turnover) was 50-60 g/day. DNL contribution to adipose-TG was 0.04 after 9 wk, representing ∼20% of newly deposited TG. Cell proliferation was 0.10-0.17 after 9 wk (half-life 240-425 days). In summary, long-term 2H2O administration to human subjects allows measurement of the dynamics of adipose tissue components. Turnover of all elements is slow, and DNL contributes ∼20% of new TG.
- triglyceride synthesis
- stable isotopes
- lipid kinetics
any disorder of adipose mass, including obesity or lipodystrophy, has an underlying kinetic basis. Synthesis and/or breakdown rates of adipose tissue lipids must be altered for changes in adipose mass to develop and be maintained. Direct measurement of the dynamics of adipose tissue components has been problematic, however, in large part because of the apparently slow turnover rate and the large pool size of adipose tissue elements (7, 11, 12, 17).
Indirect estimates of the turnover rate of adipose tissue fatty acids (FA) in humans were made by Hirsch (11) in the 1960s. Hirsch found that adipose tissue FA composition generally reflects the FA composition of diet and, moreover, that the time required for adipose tissue FA composition to change in response to a change in dietary FA composition was in the range of 6 mo (11). By inference, the residence time of adipose tissue triglyceride (TG) was interpreted to be ∼6 mo, and de novo lipogenesis (DNL) was estimated to make a minor contribution to stored fat. The rate of appearance (Ra) of plasma glycerol also provides an indirect estimate of the replacement rate of adipose tissue lipids. Whole body lipolysis, based on Ra glycerol, is ∼1 mg·kg-1·min-1 (7, 17). This represents ∼100 g fat/day in a nonobese 70-kg subject compared with adipose tissue mass of ∼15 kg. Fractional replacement of adipose-TG should therefore be ∼0.6%/day, or a residence time of ∼170 days (assuming a homogeneous and randomly replaced TG pool in whole body adipose tissue). These indirect estimates are not definitive, however.
The definitive approach to adipose tissue dynamics must involve direct measurement; this, in turn, requires metabolic labeling. A major limitation of metabolic labeling for adipose tissue is that it is difficult to administer a labeled precursor for a long enough time to achieve detectable incorporation into adipose-TG, if turnover is indeed on the order of 6 mo. Also, the optimal metabolic precursor for labeling adipose-TG has not been established. Glycerol is not utilized effectively by adipocytes (25), and different FA may have different turnover rates in adipose-TG, in addition to being difficult to administer in vivo. Glucose, although the precursor for the α-glycerol phosphate in adipocytes that is used in the synthesis of TG (25), is an inefficient label for this purpose and is not practical for long-term labeling studies.
A related question concerns the metabolic source of FA in adipose tissue TG. Stored fat may come from diet or DNL. Indirect estimates have suggested a small quantitative contribution from DNL in humans (2, 8, 10, 11, 14, 30). Direct measurement of hepatic DNL has been possible by isolation of plasma very low density lipoprotein (VLDL)-TG, which are secreted from the liver. Labeling studies have shown that hepatic DNL makes a quantitatively minor contribution to secreted VLDL-TG under most dietary conditions in humans (8, 10, 14, 30). Direct measurements of adipose tissue DNL have been more difficult, however, again because of the large pool size and slow turnover of adipose TG.
We (32, 33) recently developed a method for measuring the synthesis of TG in adipose tissue of rodents, based on the incorporation of 2H from 2H2O in the glycerol moiety of TGs, followed by mass spectrometric analysis and application of mass isotopomer distribution analysis (MIDA). This technique has the important practical feature that long-term labeling studies are extremely easy to perform, simply by allowing the animals to drink 2H2O-enriched water. It was also possible to measure DNL from 2H2O, using MIDA (32), and to measure DNA synthesis (and, thus, cell proliferation) in adipose tissue from 2H2O concurrently in rodents (23).
Here, we use the 2H2O labeling approach to measure TG synthesis, DNL, and cell proliferation in vivo in adipose tissue of healthy human subjects. We report that adipose TG turnover is indeed slow, with a TG half-life in the range of 6-9 mo, that cell turnover has similar kinetics, and that DNL contributes, on average, ∼20% of newly deposited adipose TG-palmitate, although with considerable interindividual variation. Portions of this work have been reported previously in abstract form (31, 33).
Healthy subjects were recruited by advertisement. Entrance criteria included weight stability for >3 mo, body mass index <28 kg/m2; absence of infectious or inflammatory conditions within the previous 3 mo; good general health, normal screening laboratory tests (chemistry profile, complete blood count), HIV-seronegativity; and the ability to give informed consent. Exclusion criteria consisted of prior history of metabolic disorder (diabetes, obesity, hyperlipidemia) or other organ system disease (liver, kidney, lung, etc.); use of medications with potential metabolic effects (glucocorticoids, β-blockers, thiazide diuretics, phenytoin, adrenergic agents, androgens, anabolic agents, estrogens, or oral contraceptives); and inability to give informed consent. A total of 19 subjects entered the study. Two groups of healthy subjects were enrolled: group 1 took the 2H2O for 9 wk (n = 9); group 2 (n = 10) took the 2H2O for 5 wk. We included subjects with a range of body fat, waist-to-hip ratios, and blood metabolite measurements (Table 1) in this initial survey study, with the intent of introducing variability for adipose tissue kinetic parameters. Clinical characteristics of these subjects are shown in Table 1. Five men and four women were studied in group 1, whereas group 2 consisted of 10 men. Body weight was measured at the beginning of the study and then every 2-3 wk. Body composition was measured by bioelectrical impedance analysis at the beginning and end of the 2H2O labeling study. Waist-to-hip ratio was measured by a single investigator (Strawford) in all subjects. Blood concentrations were measured by the San Francisco General Hospital Clinical laboratories. All studies and procedures received prior approval from the University of California Berkeley and University of California San Francisco Committees on Human Research, and subjects gave written informed consent before participating.
2H2O Labeling Protocol
2H2O was administered orally according to the protocol described previously (23, 24). In brief, the initial priming dose was carried out in the General Clinical Research Center (GCRC) of San Francisco General Hospital. Subjects received a total of 350-400 ml of 2H2O in the GCRC, given as divided doses over the course of 18-21 h (70 ml of 70% 2H2O, given every 3-4 h), to achieve ∼1.0% enrichment in the body water pool. The 2H2O was purchased from Isotec (Miamisburgh, OH) and dispensed in sterile containers. Subjects then took 50 ml of 70% 2H2O three times a day for 5 days and then 35-50 ml two times a day for the remainder of the 8- to 10-wk labeling protocol. This protocol achieves near-plateau body 2H2O enrichments (1.5-2.0%, see below) within 5-7 days in most subjects and was well tolerated. No subject complained of any significant adverse effects, although a mild, transient light-headed feeling was described by one subject. Subjects received the 2H2O as individual aliquots (35-50 ml of 70% 2H2O) in plastic vials, which were stored in the refrigerator.
Compliance with outpatient 2H2O intake was checked through weekly visits (for urine and saliva collection) and by return of vials for counting.
Collection of Body Water Samples and Blood for Monocytes
Plasma or urine samples were collected weekly in all subjects and frozen in closed containers. Blood was collected in Ficoll-Hypaque solution, and the mononuclear fraction was removed, after centrifugation. Blood monocytes were isolated as CD14+ cells by immunomagnetic beads, as described previously (21, 23).
Adipose Tissue Sampling Protocol and Isolation of Mature Adipocyte-Enriched Fraction
Adipose tissue aspiration biopsies were performed at weeks 5 and 9 of 2H2O intake, using the procedure described elsewhere (23). In brief, three subcutaneous sites were sampled at each visit [the gluteal region (buttocks), femoral region (thigh), and flank region (mid-back)]. For the fat aspiration procedure, lidocaine topical anesthetic (1%, with epinephrine 1:100,000) was used. Subcutaneous fat was aspirated with a 14-gauge needle in a 3-ml syringe and then placed in sterile tubes over ice for processing (on the same day). Tissue samples were minced with a sharp blade and then treated with type 2 collagenase (Worthington, Lakewood, NJ). One milliliter of a 2 U/μl solution was added to each adipose sample and allowed to incubate for 1 h at 37°C (29). The cell suspension was then poured slowly over a 350-μm mesh filter (Spectrum Laboratories, Rancho Dominguez, CA). Adipose cells in the filtrate were then collected and processed for lipid isolation and microscopic analysis.
Isolation of Metabolites for Mass Spectrometric Analyses
TG-glycerol and FA. Tissue samples were placed in Kontes dual glass tissue grinders (Kimble Kontes, Vineland, NJ) with 1 ml methanol-chloroform (2:1), ground until homogeneous, and then centrifuged to remove protein. This solution was extracted with 2 ml chloroform-water (1:1). The aqueous phase was discarded, and the lipid fraction was transesterified by incubation with 3 N methanolic hydrochloride (Sigma-Aldrich) at 55°C for 60 min. FA methyl esters The were separated from glycerol by Folch extraction with the modification that water rather than 5% NaCl was used for the aqueous phase. The aqueous phase containing free glycerol was then lyophilized, and the glycerol converted to glycerol triacetate by incubation with acetic anhydride-pyridine, 2:1 (32).
DNA from cells and isolation of deoxyribose from deoxyadenosine. DNA was isolated from the mature adipocyte-enriched fraction and the stromal-vascular fraction of adipose tissue and from blood monocytes, as described previously (5, 23), using QiAmp extraction kits (Qiagen). DNA was hydrolyzed to free deoxyribonucleosides through enzymatic hydrolysis (5, 24). The deoxyadenosine (dA) was purified from the hydrolysate using an SPE column (5), and the deoxyribose (dR) moiety of dA was derivatized to pentose-tetraacetate (PTA), as described elsewhere (24). The resulting PTA derivative was brought up in ethyl acetate for GC-MS analysis.
Body H2O. 2H2O enrichments in body water were measured from plasma or urine. A 15- to 20-μl sample was reacted in an evacuated GC vial with calcium carbide to produce acetylene (24). The acetylene gas was then transferred with a syringe, injected in an evacuated GC vial containing 10% bromine in carbon tetrachloride, and incubated at room temperature for 2 h to produce tetrabromoethane. Excess bromine was neutralized with 25 μl of 10% cyclohexene in carbon tetrachloride. The tetrabromoethane, containing hydrogen atoms from body H2O, was then analyzed by GC-MS.
TG-glycerol, FA, and H2O. GC-MS instruments (models 5970, 5971, or 5973, Hewlett-Packard, Palo Alto, CA) were used for measuring isotopic enrichments of glycerol, FA, and H2O.
Glycerol-triacetate was analyzed using a DB-225 fused silica column, monitoring mass-to-charge ratios (m/z) 159 and 160 (parent M0 and M1), or m/z 159, 160, and 161 (M0 to M1 and M2). Methane chemical ionization (CI) was used with selected ion monitoring. FA methyl esters were analyzed for composition by flame ionization detection and for 2H enrichment by GC-MS, as described elsewhere (7).
Tetrabromoethane was analyzed using a DB-225 fused silica column, monitoring m/z 265 and 266 [M0 and M1 masses of the 79Br79Br81Br (parent minus Br-) isotopomer]. Standard curves of known enrichment were run before and after each group of samples to calculate isotope enrichment.
DNA. The PTA samples were analyzed for incorporation of deuterium on an HP model 5973 MS with a 6890 GC and auto-sampler (Hewlett-Packard). Methane CI was used with a 30m DB-225 column under selected ion monitoring of m/z 245-246 (representing M0 and M1 mass isotopomers). Baseline, unenriched dA samples were measured concurrently, and the excess M1 in the adipose PTA samples was determined by difference (subtraction of the M1 measured in the standard from the M1 in the sample). Blood monocyte and granulocyte samples were run simultaneously and used to represent a completely or near-completely turned over tissue, to calculate fractional adipose cell replacement (23). An alternative calculation method is to use estimated asymptotic values for DNA, based on previously established observations from fully turned-over tissues (23). For dR analyzed as the PTA derivative, at body 2H2O enrichments of 1.0-2.0%, the EM1 values in fully turned over cells are ∼3.5 times the measured steady-state body 2H2O enrichments (3, 23). Because circulating monocytes and granulocytes are available by experimental measurement, these measured values were used.
GC-MS analytic procedures. For all GC-MS analyses, enriched samples were matched with baseline (unenriched) samples for abundance. The abundance range used was that which gave values within 1-2% of theoretical mass isotopomer ratios, as described elsewhere (9, 24). Only analytic runs for which baseline abundances achieved these accuracy levels and for which samples fell within this abundance range were considered acceptable for use in calculations.
Group comparisons were by ANOVA. Statistically significant differences were taken to be P < 0.05. Sources of variability in the measurements were assessed by using random effects models. These included random person effects (that reflect between-person variability), random depot effects within each person (that reflect depot-to-depot variation), and residual week-to-week effects (that reflect week-to-week variability within each depot for each subject). These models also included a fixed-week effect to account for systematic change over time. We also analyzed correlations between adipose TG kinetic parameters and standard (nonkinetic) parameters, using regression models. Pearson correlation and Spearman rank correlation coefficients were calculated for fractional TG synthesis, absolute TG synthesis, DNL and lipolysis vs. plasma insulin, glucose and TG concentrations, percent body fat, total body fat, and waist-to-hip ratios.
Models and Calculations
Use of 2H2O incorporation for measurement of TG-glycerol synthesis (all-source TG turnover), TG-FA synthesis (DNL), and DNA replication (cell proliferation) is described in detail elsewhere (23, 32). The techniques are summarized briefly as follows.
TG-glycerol synthesis from 2H2O. We have previously described the theory behind measurement of all-source TG synthesis and turnover based on incorporation from H-labeled water in C—H bonds of TG-glycerol (32, 33). C—H bonds of α-glycerol phosphate exchange with cellular water during the course of glycolytic and “glyceroneogenic” reactions (Fig. 1). If 2H2O is administered, the population of TG molecules that was synthesized from α-glycerol phosphate during the period of 2H2O enrichment will be exposed to 2H labeling, whereas preexisting (old) TG will be unlabeled. We have previously established (32) that the C—H bonds in glycerol of newly synthesized TG are ∼80% exchanged (i.e., 4 of 5 positions exchange) with 2H2O in body water in rodents, based on measurements of fully replaced TG and on combinatorial analysis (MIDA) of the labeling pattern in TG-glycerol (i.e., the proportion of double-labeled and single-labeled species). Once the number of exchanging hydrogens (n) is known, the fraction of newly synthesized TG molecules at time t of 2H2O administration can be calculated from measured body 2H2O enrichments (32) 1
where f is the fraction of newly synthesized TG molecules present, EM1 is the excess mass isotopomer abundance for M1-glycerol at time t, and is the asymptotic mass isotopomer abundance for M1-glycerol, based on n = 4 exchanging positions and the 2H2O enrichment measured (32).
The absolute synthesis rate of adipose tissue TG can be calculated from fractional TG synthesis multiplied by the adipose TG pool size (32). This calculation is based on the assumption that TG in subcutaneous adipose tissue depots reflects whole body TG kinetics. Although this assumption has not been tested in humans, results in rodent models suggest relative consistency among nonvisceral adipose depots (32, 33). The lipolytic rate of adipose TG (TG turnover) can also be calculated from label incorporation, based on the absolute TG synthesis rate combined with changes in pool size (32, 33). Under the conditions of these studies, where body weight and composition were stable during the 2H2O labeling period (i.e., change in adipose TG mass is 0), lipolysis equals absolute retained synthesis of adipose TG 2A 2B
Lipolysis as defined here refers to the absolute breakdown rate, including replacement of TG in adipose tissue. It should be noted that lipolysis as measured here could be affected by nonrandom turnover of adipose tissue TG, if there is any anatomical or biochemical selectivity of lipolysis. This method only detects TG synthesis that is retained in the tissue, so preferential cycling of newly synthesized TG could affect the representativeness of retained, labeled TG for true turnover (see discussion below).
DNL from 2H2O. Labeled water (3H2O or 2H2O) has been used for many years to measure DNL (14-16, 19). Techniques using 2H2O have generally involved use of isotope ratio-mass spectrometers for analysis, after reduction of H2O to H2. The technique employed here allows measurement of DNL with standard GC-MS instruments, by virtue of the fact that body 2H2O enrichments are in the range of 1-2% (see below). A few complications are introduced in the calculations, however, when 2H2O enrichments in this range are present, as we have noted elsewhere (32). A combinatorial model must be used, to account for the occurrence of multiply labeled species of newly synthesized FA that can be formed at these higher 2H2O enrichments (9). To apply a combinatorial model for DNL from 2H2O (i.e., to use MIDA; see Ref. 9), it is necessary to estimate the number (n) of hydrogen atoms in C—H bonds of FA that are incorporated from cellular H2O during DNL. The value of n can itself be calculated by MIDA, assuming that body 2H2O represents the precursor pool enrichment (p), as we (9) and Lee et al. (18) have discussed previously. The fractional and absolute contributions from DNL can then be calculated, using either measured 2H2O enrichment or MIDA-calculated 2H2O enrichment (32), for any nonessential FA 3
where is calculated from MIDA look-up tables, based on the calculated values of n and p in the FA (32) 4
Cell proliferation from 2H2O incorporation into DNA. The method for measuring DNA replication, and thus cell proliferation, based on the incorporation of 2H in the dR moiety of purine deoxyribonucleosides isolated from genomic DNA, has been described in detail elsewhere (20, 23). When 2H2O is used as the label, the fractional replacement of cells can be calculated by comparison with the theoretical asymptotic value in new dR, based on body 2H2O enrichments (20, 23), or, alternatively, by comparison with a completely (or almost completely) replaced tissue (23, 24). In humans, blood monocytes and granulocytes serve this function, as we have shown previously (23). The fractional replacement of cells in the mature adipocyte-enriched and stromal vascular fractions of adipose tissue were calculated as 6
where sample refers to the adipose tissue fraction and mono's refers to the monocyte fraction.
Body Water Enrichments
The 2H2O enrichment of body water was stable in these subjects over the course of the 9-wk outpatient labeling period (Fig. 2).
Adipose Tissue TG Turnover
The synthesis and replacement rates of adipose tissue TG were measured from the incorporation of 2H into TG-glycerol (32, 33). Incorporation into TG-glycerol increased over the course of 9 wk (Fig. 3A). Fractional synthesis (f) was ∼0.12-0.15 (12-15%) after 5 wk in the three depots sampled (Table 2) and 0.16-0.22 (16-22%) after 9 wk. Fractional replacement rates of adipose TG were in the range of 0.0030-0.0045 day-1 (Table 2), indicating relatively consistent values of half-life (t1/2) for adipose TG of ∼160-250 days (Fig. 3B).
In three subjects, follow-up adipose aspirates were performed after 5-8 mo to test label die-away rates (data not shown). Fractional decay rates ranged between 0.020 and 0.0046 day-1, similar to incorporation rates.
Adipose TG synthesis rate appeared to increase somewhat more slowly between weeks 5 and 9 than during the first 5 wk of label administration (Fig. 3A and Table 2). In the subjects matched for biopsies at both time points, f was 0.130 ± 0.048% (week 5) and 0.158 ± 0.057% (week 9) for gluteal depot, 0.151 ± 0.098 and 0.236 ± 0.127%, respectively, for flank, and 0.125 ± 0.042 and 0.205 ± 0.099%, respectively, for thigh (Table 2). Week 9 values were significantly higher than week 5 values (P < 0.05).
The total amount of adipose TG synthesized and retained over 5 wk (Table 3) was calculated to be 1.80 ± 0.98 kg (n = 18), with a range between 0.5 and 4.4 kg. The average rate of adipose TG synthesis (assuming the subcutaneous depots sampled here are representative of total body fat stores) is therefore ∼0.35 kg/wk, or 50 g/day.
Lipolysis rate could also be calculated on the basis of TG synthesis and body fat balance. Because these subjects were in zero fat balance at the whole body level (weight stable and no change in body composition over the 9-wk labeling period), net lipolysis equals net synthesis, or replacement, of adipose TG. The net lipolysis rate was calculated from the average value of the replacement rate constant k in the three depots sampled for each subject (Table 2), multiplied by the whole body fat pool size (average ∼15 kg in these subjects, Table 1). These values were ∼45-60 g TG/day, or ∼0.4-0.6 mg TG·kg body wt-1·min-1 (Table 3).
Contribution from DNL to Adipose Tissue TG
Incorporation into TG-FA increased in a roughly linear manner between weeks 0 and 9 (Table 4). Fractional DNL (fDNL) reached an average of 0.020 ± 0.012 in gluteal fat (n = 17), 0.023 ± 0.016 in flank fat (n = 17), and 0.025 ± 0.013 in thigh fat (n = 12) after 5 wk of 2H2O labeling (Table 4). After 9 wk of 2H2O labeling, fDNL values were 0.040 ± 0.025 in gluteal fat (n = 8), 0.041 ± 0.024 in flank fat (n = 6), and 0.041 ± 0.024 in thigh fat (n = 8). When corrected for the fraction of adipose TG that was newly deposited (i.e., correcting for TG-glycerol synthesis), the fractional contribution from DNL to newly deposited TG-palmitate was relatively constant over time and among depots within most individuals (Table 4), with an average value of ∼20%. Individuals appeared to be on their own characteristic curve, however, with consistent results within each individual over time and for all depots tested.
Interindividual variability was observed for TG synthesis and DNL (Tables 2 and 4). Ranges for fractional TG synthesis measured were from 0.03 to 0.32 replaced after 5 wk and 0.08 to 0.49 after 9 wk (Table 2). Fractional DNL contribution to palmitate (corrected for TG replacement) ranged from 0.049 to 0.435 (Table 4). No significant differences between depots were observed for any parameter. For both fractional TG synthesis and corrected DNL contribution, person-to-person variation accounted for more than one-half of the total variability (51 and 55%, respectively) and week-to-week variation accounted for most of the rest (40 and 35%, respectively). Variation between depots in the same subject at the same time point was only 9 and 10% of total variability for the two parameters, respectively. Because depot-to-depot variability within subjects was small compared with interindividual variability, measurement error appears to be acceptably small, and interindividual differences likely represent true variability among human subjects.
As an initial attempt to understand the biological basis of interindividual variability, we analyzed correlations between adipose lipid kinetic parameters and standard (nonkinetic) measures using regression models. Fractional synthesis of TG, absolute synthesis of TG, DNL contribution, and lipolysis rate were analyzed against the percent body fat, total body fat, waist-to-hip ratio, and plasma glucose, insulin, and TG concentrations [average values of 79 ± 8 mg/dl (range 64-90), 15.0 ± 5.5 μU (range 7.4-27.0), and 104 ± 52 mg/dl (range 65-223)]. Several correlations were significant and had r2 values >0.10 (Fig. 4). These correlations included fractional TG synthesis vs. waist-to-hip ratio (r2 = 0.17), DNL vs. plasma glucose concentration (r2 = 0.11), absolute TG synthesis vs. percent body fat (r2 = 0.21), absolute TG synthesis vs. serum insulin concentration (r2 = 0.20), fractional TG synthesis vs. plasma TG concentration (Spearman rank correlation 0.56), lipolysis vs. waist-to-hip ratio (r2 = 0.34), and lipolysis vs. percent body fat (r2 = 0.25).
Cell Proliferation (DNA Synthesis) in Adipose Tissue
The mature adipocyte-enriched fraction contains >95% large, nucleated, lipid-rich cells (i.e., mature adipocytes) by microscopic analysis of human or rodent adipose tissue aspirates prepared by the method described here (D. Cesar, C. Loe, and M. Hellerstein, unpublished observations). Measurement of DNA synthesis, representing cell proliferation (5, 23, 24), in the mature adipocyte-enriched (stromal-vascular depleted) fraction revealed progressively increasing incorporation between weeks 5 and 9 (Fig. 5). After 5 wk of 2H2O intake, fractional replacement rates were 0.066 ± 0.035 (gluteal, n = 9), 0.050 ± 0.029 (flank, n = 8), and 0.061 ± 0.048 (thigh, n = 7). After 9 wk of 2H2O, fractional replacement rates were 0.167 ± 0.141 (gluteal, n = 9), 0.093 ± 0.108 (flank, n = 7), and 0.103 ± 0.056 (thigh, n = 7). Fractional replacement rate constants were therefore calculated to be in the range of 0.0016-0.0029 day-1 (i.e., 0.16-0.29%/day, t½ = 240-425 days). The fractional replacement of cells in the stromalvascular fraction was consistently and significantly higher than in the mature adipocyte-enriched fraction (Fig. 5). After 5 wk of 2H2O labeling, fractional replacement was 0.192 ± 0.166 for the gluteal stromal-vascular fraction (n = 5), 0.194 ± 0.120 in flank (n = 4), and 0.195 ± 0.156 in thigh (n = 2). After 9 wk of labeling, stromal-vascular replacement fractions were 0.236 ± 0.104 in gluteal fat (n = 6), 0.291 ± 0.075 in flank (n = 5), and 0.353 ± 0.227 in thigh (n = 6). The number of subjects with cell proliferation data (DNA kinetics) is lower than for lipid kinetics because cells have to be isolated and separated fresh, so any problems with yield are final, in addition to the higher content of lipids vs. cells in adipose tissue.
The approach described here allows concurrent measurement of adipose tissue TG synthesis, net lipolysis (TG breakdown), contribution from the DNL pathway, and proliferation rate of cells, integrated over long periods of time (i.e., weeks or months) in human subjects. Moreover, because outpatient intake of heavy water (2H2O) is simple, easy to comply with, and relatively inexpensive, this approach has a number of practical advantages over alternative methods for estimating dynamics of adipose tissue components in humans.
Measurement of TG replacement (turnover) rates from the incorporation of 2H in the glycerol moiety of acylglycerides is a new approach in humans. We have used this technique extensively in rodents (31-33), and Brunengraber et al. (2a) also described a similar approach in rodents, but this is the first application of the method in human subjects. Turnover rates of adipose TG calculated from this 2H2O/TG-glycerol method were similar to previous estimates from indirect techniques (11, 12), i.e., t1/2 of ∼6 mo. Interestingly, the turnover rate of cells in adipose tissue was of a similar magnitude. If subcutaneous adipose-TG is taken to be representative of TG turnover in the whole body (clearly an oversimplification; see Ref. 27), the daily TG deposition and breakdown rate (= k × whole body TG pool size) is ∼50-60 g/day compared with an average dietary fat intake of ∼100 g/day in the Western diet. These measurements suggest that the fraction of dietary TG that undergoes oxidation without prior storage in adipose TG is roughly equal to the fraction that is stored in adipose TG.
Previous kinetic studies in rodents have clearly demonstrated that visceral fat stores are more lipolytically active than subcutaneous fat (32). The inclusion of visceral fat samples in our measurements would therefore probably increase calculated whole body TG synthesis and lipolysis rates. Because the preponderance of body fat in nonobese subjects is in subcutaneous adipose stores, the results presented here are unlikely to be significantly altered by the absence of visceral depots, although perhaps somewhat underestimated. In subjects with abdominal obesity, however, extrapolation to whole body TG turnover rates may not be valid.
An interesting variation on this method is that label incorporation in adipose TG can be used as a measure of TG breakdown (lipolysis). Label incorporation has long been used to measure catabolic rates in fields such as protein turnover (34) and cell turnover (6), for example, but not traditionally for adipose TG breakdown. Lipolysis has generally been measured by the dilution of blood metabolites (e.g., glycerol or free FA; see Refs. 7 and 17). It should be recognized that the approach described here does not measure the identical process as plasma glycerol flux, because any preferential recycling of a subfraction of newly deposited adipose-TG (i.e., any nonrandom turnover of the adipose TG pool) will result in lower estimates of TG breakdown based on net label retention over the long-term than by short-term dilution of plasma metabolites. We are aware of no evidence for preferential recycling of a subpopulation of newly deposited TG in adipose tissue. In any event, the values for net lipolysis (∼0.5-0.75 mg TG·kg body wt-1·min-1) were reasonably similar to Ra glycerol values reported elsewhere in healthy, nonobese humans (0.1-0.15 mg glycerol·kg body wt-1·min-1 or ∼1.0-1.5 mg TG·kg body wt-1·min-1; see Refs. 7 and 17), although lower (as expected). Direct comparison of these two approaches will be of interest in future studies. The 2H2O approach represents an alternative, non-infusion-based approach to measurement of adipose-TG turnover that may have advantages in certain settings.
Estimates of the DNL contribution to adipose-TG were also interesting. The average value of ∼0.20 (20%) DNL contribution to newly deposited adipose TG-palmitate is higher than our previous estimates of hepatic DNL, based on shorter-term label incorporation in circulating VLDL-TG-palmitate (2, 8, 10, 30). We believe that it is likely, but we cannot be certain, that this difference reflects adipose DNL. Extending the period of administration of [13C]acetate or 2H2O from the usual 12-16 h up to 60-80 h in humans does increase the apparent DNL contribution to VLDL-palmitate secreted from the liver (13), presumably reflecting the time needed to fully label hepatic lipid storage pools that contribute to VLDL-TG synthesis. The increase in estimated DNL contribution that results from prolonging label infusions is modest, however (∼1.5-fold, to ∼10%), and the hepatic DNL contribution almost never reaches 20% under conditions of a typical high-fat Western diet in healthy subjects (8, 10). We therefore believe that the consistently higher values of DNL in adipose-TG after long-term labeling, compared with VLDL-TG after shorter-term labeling, represent more than an artifact of labeling time. Most likely, this difference represents evidence for DNL input in human adipose tissue. If this inference can be tested more directly, it would be important for our understanding of human macronutrient balances. Some previous work (1, 26) has indirectly suggested that adipose DNL is capable of substantial induction in humans under conditions of massive carbohydrate overfeeding, but this proposed explanation was not tested experimentally. It is important to recognize, however, that DNL still contributed a small fraction of total adipose TG storage in the free-living subjects studied here, even with the higher estimates here from the 2H2O method. Nonessential FA represent about one-half of stored adipose FA. If 20% of nonessential FA come from DNL, ∼10% of total stored FA derive from DNL. This contrasts with rodents, where long-term 2H2O administration results in up to 70% of palmitate deriving from the DNL pathway in animals on low-fat diets (22, 32, 33). Whether the difference between our findings with 2H2O labeling in humans compared with rodents reflects species differences in adipose tissue lipogenic capacity, or differences in diet, activity, hormones, or other factors is an important question to pursue that is at present unresolved.
The observation of substantial interindividual variability for adipose tissue lipid kinetics and biosynthetic pathways (Tables 2 and 4) may prove a fertile area for follow-up investigation. Statistical analyses showed much greater variability among individuals (>50% of total variability) than among depots in the same subject at the same time point (<10% of total variability). Measurement or technical error is therefore small, and interindividual differences likely represent true variability among subjects. The biological sources of interindividual differences remain uncertain but potentially interesting. We documented modest, although significant, correlations between adipose lipid biosynthetic parameters and some standard (nonkinetic) measures (Fig. 4). None of the observed relationships was striking, however. Other factors not measured here, such as diet, activity, genetics, etc., may explain most of the variability among individuals that we observed. Experimental evaluation of potential modulating factors should be possible using the labeling approach described here.
The cell proliferation kinetics in adipose tissue (Fig. 5) were similar to, although slower than, the kinetics of adipose TG (Table 2). The mature adipocyte-enriched (stromal-vascular depleted) fraction that we isolated from adipocyte tissue contains >95% mature adipocytes, based on microscopic analysis. It is possible that rare islands of vascular cells might contribute nonadipocyte DNA out of proportion to microscopic purity. However, recent kinetic measurements of mature adipocyte fractions isolated to even higher purity from rodents (>98% purity) in our laboratory gave nearly identical results as the method used here (S. Turner, C. Loe, and M. K. Hellerstein, unpublished observations). Accordingly, we believe that a significant contaminating effect of nonadipocytes is unlikely. The stromal-vascular cell fraction did have a consistently higher turnover rate than the mature adipocyte-enriched fraction (Fig. 5). It is therefore possible that contamination by stromal-vascular cellular elements could contribute to some of the observed proliferation in the mature adipocyte-enriched fraction, although this effect is likely to be minor, for the reasons just noted. Moreover, accrual of adipose tissue as an organ has been shown to involve coordinate induction of stromal-vascular elements and adipocytes (4). Measurement of proliferation of nonadipocyte components of adipose tissue is therefore of physiological interest in its own right. The cell kinetic results may be somewhat less certain than the TG kinetics, however.
A technical caveat is worth mentioning. The 2H2O labeling protocols were not carried out for a long enough time to determine whether the asymptotic value of adipose-TG turnover being approached was 100% (32). The lack of a plateau value does not affect the calculation of fractional replacement. In some individuals, there appeared to be a pseudo-plateau for adipose TG fractional synthesis (Fig. 3A and Table 2). These observations could represent an anatomical or biochemical subpopulation of adipose TG that is preferentially turning over (discussed above). More detailed characterization of longitudinal incorporation kinetics in adipose TG will be required to answer this question definitively (i.e., to determine whether a subpopulation of the adipose TG pool is preferentially turning over).
In summary, the dynamics of adipose tissue TG, DNL, and (with some uncertainty) cellular elements are measurable concurrently in human subjects through use of a single tracer (2H2O). Turnover of both lipids and cells is slow (t1/2 6 mo or longer) in adipose tissue of humans, consistent with previous indirect estimates. The DNL contribution to newly deposited adipose TG is ∼20%, with considerable interindividual variability. Factors influencing the dynamics of adipose tissue components can now be investigated directly in humans through use of this simple labeling approach.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by American Physiological Society