A Multicompartmental Model of In Vivo Adipose Tissue Glycerol Kinetics and Capillary Permeability in Lean and Obese Humans

Twenty-four subjects participated in this study (Table 1). All subjects were weight stable for at least 2 months before the study and were considered to be in good health after completing a comprehensive medical evaluation, including history and physical examination, blood tests, and electrocardiogram. Six subjects were obese (BMI >30 kg/m²). In 15 subjects (12 female), body composition was measured by using dual-energy X-ray absorptiometry (Lunar Instruments, Madison, WI) within 2 weeks before the study. None of the subjects were taking regular medication, and premenstrual female subjects were studied during the follicular phase of their cycle. All obese subjects had normal glucose tolerance confirmed by a standard oral glucose tolerance test. The study was approved by the Human Studies Committee and the Clinical Research Center Scientific Advisory Board, and all subjects gave informed written consent. Other aspects of these studies have been previously reported (24).

Subjects were admitted to the Clinical Research Center in the afternoon before the study. At 1800, subjects ingested a meal containing 12 kcal/kg body wt for lean subjects and 12 kcal/kg adjusted body wt for obese subjects (adjusted body weight = ideal body weight [IBW] + [actual body weight − IBW] × 0.25). At 2000, subjects ingested a defined liquid formula snack containing 250 kcal, 40 g carbohydrate, 6.1 g fat, and 8.8 g protein (Ensure; Ross Laboratories, Columbus, OH). After this snack, all subjects fasted until completion of the study the following day.

The following morning, 20-gauge catheters were inserted into a forearm vein for isotope infusion and into a radial artery for arterial blood sampling. A superficial abdominal vein was cannulated with a 10- to 20-cm, 22-gauge polyurethane catheter (Hydrocath; Viggo-Spectramed, Oxnard, CA) (7,9,12,24). Blood obtained from this site represents effluent from adipose tissue and overlying skin. All vascular catheters were kept patent by continuous saline infusion.

Three or four microdialysis probes (CMA, Acton, MA) were placed percutaneously without anesthesia into subcutaneous abdominal adipose tissue. Each probe consisted of dialysis tubing (10 × 0.5 mm, 20,000 MW cutoff) and was perfused overnight with Ringer’s lactate solution (0.3 μl/min) before insertion to ensure the elimination of any glycerol from the catheter itself. Perfusion of the probe permits the equilibration of adipose tissue interstitial glycerol with perfusate, which can be collected. All probes were placed within 10 cm of the midline and were at least 3 cm apart from each other. Each was continuously perfused with lactated Ringer’s solution at a rate of 0.1 μl/min by using a syringe infusion pump (Harvard Apparatus, South Natick, MA). Both venous and microdialysis catheters and the ¹³³Xe depot were positioned so as to sample the more superficial subcutaneous tissue as described by Enevoldsen et al. (25).

Subjects remained supine throughout the study, and room temperature was kept constant at 23°C during the entire study. Baseline arterial and abdominal venous blood samples were obtained 60 min after catheters were placed. Adipose tissue interstitial fluid samples were collected 60 min after probe insertion. The first 60-min fraction of the dialysate effluent was discarded to eliminate the influence of the initial trauma caused by probe insertion on glycerol measurements. Previous studies (11,12) have shown a transient rise in interstitial fluid ATP, an index of tissue damage, during the first 15 min after probe insertion. After baseline, samples were collected and a primed (3.6 μmol/kg)-continuous (0.24 μmol · kg⁻¹ · min⁻¹) infusion of 1,1,2,3,3-[²H₅]glycerol (Tracer Technologies, Somerville, MA) was started and maintained for 120 min using a syringe infusion pump (Harvard Apparatus, South Natick, MA). Arterial and abdominal venous blood samples were taken at 45, 60, 75, 90, 105, 110, 115, and 120 min. Dialysate from the microdialysis probes was collected for 60 min, between 60 and 120 min of isotope infusion, and the fluid from all probes were pooled to ensure recovery of adequate amounts of glycerol to permit accurate measurement of isotopic enrichment.

Abdominal subcutaneous adipose tissue blood flow (ATBF) was evaluated using the ¹³³Xe washout technique (7,11,12,19,25). Immediately after starting the isotope infusion, 120–150 mCi of ¹³³Xe dissolved in 0.1 ml saline was injected over 60 s into the subcutaneous abdominal adipose tissue space. The decline in ¹³³Xe activity was monitored continuously from 60 to 120 min after injection with a sodium iodide scintillation detector (Canberra Industries, Meriden, CT) placed ∼40 cm from the ¹³³Xe depot and coupled with a multichannel analyzer (ND 600; Schaumburg, IL) set to measure the 81 keV ¹³³Xe photopeak.

Arterial and venous plasma triglyceride concentrations were measured enzymatically (26). Arterial and venous blood concentrations of glycerol were determined enzymatically with an automated analyzer (Technicon, Tarrytown, NY). Isotopic enrichment (tracer-to-tracee ratio [TTR]) of glycerol in plasma and in microdialysate fluid was determined by gas chromatography–mass spectrometry by using an MSD 5971 system (Hewlett-Packard, Palo Alto, CA) with a 12 m × 0.2–mm HP-1 fused silica capillary column (Hewlett-Packard). Plasma was deproteinized with barium hydroxide and zinc sulfate and then cetrifuged to isolate the protein-free supernatant. The microdialysate samples and the supernatant of the plasma samples were passed through stacked cation (Dowex AG-50W-X8) and anion (Dowex AG-1-X8) exchange columns. A trimethylsilyl derivative of glycerol was formed and injected into the gas chromatography–mass spectrometer. Ions, produced by electron impact ionization, were selectively monitored at mass-to-charge ratios 205.1, 206.1, and 208.1.

ATBF was calculated from ¹³³Xe clearance as previously described (7,19,26), assuming an adipose tissue–to–blood partition coefficient for xenon of 10 ml/100 g for all subjects (17–19,27). Whole-body glycerol rate of appearance, R_a(systemic) (μmol · kg⁻¹ · min⁻¹), in blood was calculated by using Steele’s equation for steady-state conditions modified for use with stable isotopes (28,29): R_a(systemic) = I/TTR_art, where I is the isotope infusion rate in μmol · kg⁻¹ · min⁻¹ and TTR_art is the TTR of glycerol in arterial plasma at isotopic equilibrium.

Regional subcutaneous abdominal adipose tissue net glycerol release rate [R_a(local)], in ng · l00 g adipose tissue⁻¹ · min⁻¹, was calculated by using standard principles of arteriovenous balance and blood glycerol concentrations (29): R_a(local) = ATBF × (Gly_vein − Gly_art), where ATBF is the rate of subcutaneous ATBF in ml · 100 g adipose tissue⁻¹ · min⁻¹ and Gly_art and Gly_vein are arterial and venous blood glycerol concentrations in μmol/l.

R_a(local) (μmol · 100 g tissue⁻¹ · min⁻¹) was also calculated by using standard arteriovenous balance methodology in conjunction with isotope tracer enrichment data (29): R_a(local) = ATBF × Gly_art × ([TTR_art/TTR_vein] − 1).

We adapted the approach of Biolo et al. (23) to produce a four-compartment model for adipose tissue glycerol metabolism (Fig. 1) that contains several components. First, glycerol enters adipose tissue interstitial fluid after it is released by hormone-sensitive lipase–mediated lipolysis of adipocyte triglyceride. Interstitial glycerol can enter the venous blood compartment (F_VI) by a process that depends on the permeability of capillary endothelium. We assumed that permeability for glycerol efflux is the same as that of glycerol influx in adipose tissue. Second, arterial delivery of glycerol to adipose tissue [F_(in)] is determined by the product of Gly_art and blood flow. Glycerol present in the arteriolar compartment can pass directly into the vein by functional shunting (F_VA) or into adipose tissue interstitium (F_IA). Third, glycerol released into adipose tissue venous blood by LPL action on circulating triglycerides (F_LPL) was considered to originate on the intraluminal side of the endothelial barrier. This glycerol can enter the interstitial compartment (F_IL) or be delivered to the systemic circulation by directly entering veins that drain adipose tissue (F_VL).

Our model is not able to determine the exact proportion of LPL-derived glycerol directed to the vein or adipose tissue interstitium. However, we can examine two boundary conditions: 1) when all LPL-derived glycerol enters the interstitium and none enters the vein directly (boundary condition P) and 2) when all LPL-derived glycerol enters the vein directly and none enters the interstitium (boundary condition Q). For any individual subject, the true value of each model parameter estimate lies between the two boundary conditions.

By definition, F_VL = 0 and F_IL = F_(LPL). The relationship between arterial, abdominal venous, and interstitial glycerol is calculated as (20,21) F_(out) = F_VI + F_VA;

The fluxes F_(out), F_VI, and F_VA between compartments are indicated in Fig. 1. TTR_art, TTR_vein, and TTR_interstitial are glycerol TTR in arterial blood, venous blood, and interstitial fluid, respectively. This equation can be rearranged to (TTR_art − TTR_interstitial) × F_VI = (TTR_art − TTR_vein)F_(out);

By definition, F_VL = 0 and F_IL = F_(LPL). The relationships between arterial, abdominal venous, and interstitial glycerol are calculated as F_VA + F_VI + F_LPL = F_(out);

Therefore,

F_(out) is calculated as the product of glycerol concentration in the vein, and ATBF and F_LPL are the product of plasma flow and the arteriovenous difference in triglyceride concentration. Summers et al. (12) considered that glycerol released by LPL would not enter the interstitium in significant amounts, i.e., boundary condition Q is physiologically more plausible.

For both boundary conditions, we used the parameters from our model in conjunction with single-pass unidirectional extraction (E) principles (21) to estimate the capillary diffusion capacity (ps), which is a measure of the capillary surface area and the permeability of the capillaries to glycerol. The single-pass unidirectional extraction, which is the amount of glycerol arriving in the arterial blood that is transported into the tissue, is estimated by the model E = F_IA/F_(in), where F_IA = F_(in) − F_VA.

However, E is also fundamentally related to capillary diffusion capacity (21): 1 − E = e^(−ps/ATBF), where ps is the capillary diffusion capacity expressed in ml · 100 g tissue⁻¹ · min⁻¹. These equations can be combined to the following: ps = −ATBF 1og_n (l − E).

Having calculated ps and E allows determination of interstitial glycerol concentration Gly_interstitial, because Gly_interstitial = (Gly_vein − Gly_art × [1 − E])/E;

Data from previous studies allowed us to expect additional constraints on the model. First, the data from previous studies (7,19) demonstrated no metabolism of labeled glycerol by adipose tissue. Therefore, the model does not need to include uptake of plasma-free glycerol into adipose tissue. Second, we assumed that the incorporation of labeled glycerol into triglyceride-rich lipoproteins was negligible during the short duration of the tracer infusion study. In similar studies, we found that isotopic enrichment of glycerol within triglyceride-rich lipoprotein was <3% plasma glycerol enrichment within this time of infusion (B.W.P., S.K., unpublished observations).

The glycerol concentration and enrichment data were normally distributed. To assess for steady-state conditions, we undertook an ANOVA of the plasma concentrations and enrichments, seeking an effect of time. Some model parameters were not normally distributed; therefore, nonparametric statistics were used when possible. Data are expressed as median (interquartile range), and Wilcoxon’s and Spearman’s (R_s) tests were used to analyze the data. A P value of ≤0.05 was considered to be statistically significant.

The model parameters for glycerol kinetics are expressed as the median of the parameters from the model run on the 24 subjects’ observational data (i.e., median of the models), which is shown in Tables 2 and 3. However, as an example of what individual data looked like, Figs. 2A and B show the parameter values obtained by entering the median observational data (concentrations and specific enrichments) from 24 subjects into the model (i.e., the model of the median data).

The measured glycerol TTR in artery, abdominal vein, and abdominal interstitial fluid and concentration in artery and abdominal vein are shown in Fig. 3. The time course for circulating glycerol concentration and TTR demonstrate the presence of physiologic and isotopic steady-state conditions; ANOVA showed no significant changes with time. Arterial blood glycerol concentration (median [interquartile range] was 69.1 μmol/l [56.5–85.5], significantly lower than abdominal venous blood glycerol (230 μmol/l [210–268], P < 0.01). The model-derived values (using both boundary conditions) of adipose tissue interstitial glycerol concentration (Table 2) were greater than measured abdominal venous blood glycerol concentration (P < 0.001 for both).

Glycerol TTR was consistently greater in arterial plasma (8.34% [7.44–10.1]) than abdominal venous plasma (2.34% [1.71–2.69]) (P < 0.001), which was consistently greater than glycerol TTR in abdominal interstitial fluid (1.44% [1.11–1.79]) (P < 0.001 for all comparisons), indicating flux of glycerol from adipocytes to interstitial fluid to the vascular compartment (Fig. 3). There was no significant uptake of glycerol tracer across adipose tissue; labeled blood glycerol content was 2.4% (−3.7 to 5.0) lower in abdominal venous than arterial samples, not statistically significantly different from zero.

Triglyceride concentrations were consistently lower in abdominal venous plasma (731 μmol/l [647–1,133]) than arterial plasma (784 μmol/l [678–1,188], P < 0.001).

Median abdominal subcutaneous ATBF was 3.07 ml · 100 g tissue⁻¹ · min⁻¹ (2.13–4.06). ATBF correlated with ps (for both boundary conditions (R_s ≥ 0.55, P < 0.01) (Fig. 4), and blood flow was lower in obese than lean subjects (e.g., R_s with BMI = −0.625, P < 0.001).

Systemic glycerol R_a was 2.11 μmol · kg⁻¹ · min⁻¹ (1.71–2.31). Local adipose tissue glycerol release rate calculated by arteriovenous balance was 483 nmol · 100 g tissue⁻¹ · min⁻¹ (339–630). Local glycerol release calculated from isotope balance was not significantly different. Glycerol release by LPL action contributed 18.3% (15.0–21.2) of the total local glycerol release. The proportion of local glycerol release that came from LPL action decreased with BMI and percentage IBW (R_s > 0.46, P < 0.05 for both).

The results of the compartmental modeling of boundary conditions P and Q are shown in Table 2. Figs. 2A and B give an example of the parameter estimates from a single dataset. Table 3 shows parameters for lean (BMI <25 kg/m²) and obese (BMI >30 kg/m²) subjects.

The proportion of arterial glycerol that entered the interstitial space, F_ia/F_in, was 60.1% (47.8–81.0) in boundary condition P. This proportion also represents the single-pass unidirectional extraction value. The boundary condition P values for capillary diffusion capacity (ps) and interstitial glycerol concentration are shown in Table 2. Neither ps nor interstitial glycerol concentration was significantly different between male and female subjects. Age was related to ps (R_s = 0.43, P < 0.05) but not to interstitial glycerol concentration. There was an inverse relationship between ps value and adiposity (correlation ps with BMI R_s = −0.49, ps with percentage IBW R_s = −0.45, ps with percent body fat R_s = −0.54, P < 0.05 for all). There were positive correlations between interstitial glycerol concentration and adiposity (correlation with BMI R_s = 0.50, with percentage IBW R_s = 0.52, with percent body fat R_s = 0.52, P < 0.05 for all).

In boundary condition Q, by definition, all glycerol released by LPL action entered the venous blood without traversing the interstitial space. Each of the glycerol model parameters, F_IA, F_VA, and F_VI, calculated for boundary condition Q were different from the parameters calculated for boundary condition P in all subjects (P < 0.001 for all model parameters).

The proportion of arterial glycerol that entered the interstitial space, F_ia/F_in, was 52.8% (40.8–65.3), which was consistently less than that seen in boundary condition P (P < 0.001). This proportion also represents the single-pass unidirectional extraction value.

The calculated values for capillary diffusion capacity (ps) and interstitial glycerol concentration in boundary condition Q are shown in Table 2. Neither ps nor interstitial glycerol concentration was significantly different between male and female subjects. Permeability surface area (ps) was related to age (R_s = 0.42, P < 0.05), but interstitial glycerol concentration was not. There was significance between ps value and adiposity (correlation with BMI R_s = −0.58, with percentage IBW R_s = −0.49, with percent body fat R_s = −0.68, P < 0.05 for all). There were positive correlations between interstitial glycerol concentration and adiposity (correlation with BMI R_s = 0.41, with percentage IBW R_s = 0.41, with percent body fat R_s = 0.52, P < 0.05 for all).

In the present study, we adopted a novel approach to study postabsorptive regional abdominal subcutaneous adipose tissue glycerol kinetics in human subjects. Our data demonstrate that glycerol released from subcutaneous adipocytes travels along a concentration gradient from adipose tissue interstitial space to local veins before entering the systemic circulation. Of the glycerol appearing in the venous drainage from adipose tissue (median 732 nmol · 100 g tissue⁻¹ · min⁻¹), 52% (383 nmol · 100 g tissue⁻¹ · min⁻¹) was derived from lipolysis of adipocyte triglycerides and 14% (99 nmol · 100 g tissue⁻¹ · min⁻¹) from lipolysis of circulating plasma triglycerides. Between 8 and 11% (58–82 nmol · 100 g tissue⁻¹ · min⁻¹, for boundary condition P and Q, respectively) of venous glycerol had been functionally shunted directly from arterial to venous blood without uptake into the interstitium. The remainder of glycerol that appeared in adipose tissue venous drainage was comprised of arterial glycerol that passed into the interstitium before reappearing in the local venous circulation.

Our approach also allowed calculation of the ps index of the capillary perfusion capacity for glycerol in subcutaneous abdominal human adipose tissue. Unfortunately, ps is difficult to measure, and other groups have suggested values ranging from 3.0 to 5.0 ml · 100 g tissue⁻¹ · min⁻¹ (11,17–19,30), based in vitro methods (21,31–34). Our calculated values for ps were based on measurements made in vivo and were generally lower than ps values assumed in studies (11,15,17–19,30). In our study, ps values declined with increasing adiposity, consistent with the histological changes in adipose tissue observed in obese subjects (35,36). Our data suggest that studies of regional lipolytic activity that depend on interstitial glycerol concentration measurements should assume different ps values in lean and obese subjects. It is not known whether different adipose tissue beds have different capillary diffusion capacities, although data from a recent study (22) have shown that ps in human muscle changes in response to physiological stimuli. In our study, variation in ps was significantly related to obesity and age rather than sex, but further work would be necessary to examine whether other factors (e.g., insulin resistance or abnormal LPL activity) are more closely linked to variation in ps. In our study, ps was also related to ATBF, which itself is known to be reduced in obesity (17,18,27). It is likely that both ATBF and ps are reduced by structural alterations that occur in adipose tissue with increased adiposity (i.e., increased distance between capillary and adipocyte, and decreased capillary density); however, the association with blood flow does make it difficult to resolve the independent effects on ps of obesity and of low blood flow.

The ps for glycerol is probably related to the ps for other substances within the same tissue, because glycerol appears to diffuse in and out of tissues without any active transport processes. The methods we have used would allow future exploration of the factors that change the capillary diffusion capacity of human adipose tissue. The consequence of assuming an erroneous value for ps on local glycerol release rate is underscored in Table 3. Assuming a ps value of 5, rather than using the values calculated from our observations, caused an artifactual increase in local glycerol release rate but did not change the comparison between lean and obese subjects.

The validity of our calculations of ps depends upon the limitations and assumptions of our methods. These assumptions include the following. 1) There was no metabolism of glycerol by adipose tissue, 2) a true steady state was achieved in all sampling sites, 3) microdialysate samples were in equilibrium with interstitial glycerol with no isotope effect, 4) samples obtained by microdialysis were representative of adipose tissue interstitial fluid that was drained by the abdominal vein cannula, and 5) LPL-induced clearance of plasma triglycerides produces glycerol in equimolar amounts with little generation of mono- or diglycerides (37). We also assumed that there was no significant glycerol tracer recycling from adipocyte or plasma triglycerides. The possibility of tracer recycling is unlikely because the adipocyte triglyceride pool is very large, and any labeled glycerol that found its way there would be infinitely diluted. Moreover, there was little incorporation of labeled glycerol into plasma triglyceride during the course of this study.

In the present study, there was no significant glycerol uptake by adipose tissue, which is consistent with previous studies (7,19) involving labeled glycerol, and the absence of significant glycerol kinase activity in adipose tissue (38). In contrast, Lee et al. (39) reported acylation of glycerol in several tissues and Kurpad et al. (40) reported significant glycerol-isotope uptake by subcutaneous adipose tissue. However, the Kurpad et al. study may not have achieved steady-state conditions, glycerol balance being assessed after only 60 min of labeled glycerol infusion. In the present study, labeled glycerol was infused for 120 min, and the time-course data (Fig. 3 and ANOVA) for plasma glycerol concentration and TTR suggest that steady-state conditions were achieved. Although the other assumptions cannot be specifically tested, they have been accepted in previous studies.

Many results seen in this study were similar to previous values. Thus, whole-body glycerol R_a was similar to previous reports (5–7), as were arterial and abdominal venous blood glycerol concentrations (7,11,12,27). The values of calculated interstitial glycerol concentration in the current study are similar to those reported previously and, as expected, show a direct correlation with increasing adiposity (17–18). However, our calculated interstitial glycerol concentrations were higher than measured abdominal venous blood glycerol. In contrast, in a previous study (12) that measured interstitial glycerol concentration directly by microdialysis sampling, the value for adipose tissue interstitial glycerol concentration was lower than venous plasma glycerol concentration. The reason for this discrepancy between studies may be related to the technical difficulty in accurately measuring adipose tissue interstitial glycerol concentration directly, which have at times (e.g., Maggs et al. [41]) yielded results markedly discrepant with other studies (13) and the values we report. Early studies (14–17) that measured adipose tissue interstitial glycerol concentration by using microdialysis involved a system that had poor glycerol recovery and required time-consuming calibrations. Our calculated values for adipose tissue interstitial glycerol concentration did not require probe calibration or an estimate of probe glycerol recovery. Current probes have >90% glycerol recovery when the perfusion rate is very slow (42).

The action of LPL on intravascular circulating triglycerides can release glycerol into the circulation. This study and previous studies (7,12,27,43) show significant clearance of circulating triglyceride by adipose tissue. In this study, the proportion of local glycerol release from LPL action declined significantly with obesity in accordance with previous comparisons of lean and obese subjects (27,44).

Some LPL-generated glycerol could diffuse into the interstitium before later release into venous plasma, while another fraction of LPL-generated glycerol would go to the venous blood without traversing the interstitium. Because we did not know the fraction of the LPL glycerol following these two paths, we considered the boundary conditions where 100% of the LPL-generated glycerol goes into the vein directly and none goes into the interstitium (boundary condition Q) or the converse (boundary condition P). The true situation lies between these two extremes, but one can draw inferences from the similarities of the two boundary conditions. LPL action on triglyceride-rich lipoproteins occurs within the capillary lumen (44) so it is plausible that most LPL-generated glycerol is “swept away” directly into the venous blood rather than accessing the interstitial fluid glycerol pool, as argued by Summers et al. (12). However, some LPL-generated glycerol might be taken up into the interstitial space without mixing with venous blood. Unstirred water layers exist at the boundary of most cell surfaces (45) and such local conditions could theoretically allow LPL-generated glycerol to cross the capillary barrier and enter the interstitial space.

The findings of the present study may not represent glycerol kinetics in other fat depots. In vitro studies of isolated adipocytes demonstrate regional differences in lipolytic activity and hormonal regulation of lipolysis. For example, abdominal adipose tissue is more lipolytic, more sensitive to β-adrenergic stimulation, and more resistant to insulin-mediated antilipolysis than gluteal or femoral fat depots (46–48). Such heterogeneity of adipose tissue emphasizes that results from a single depot such as the one studied here should be extrapolated to whole-body fat metabolism with caution.

Our results also confirm some limitations of using systemic glycerol R_a as a measure of whole-body adipose tissue lipolysis. We found that glycerol released during lipolysis of plasma triglycerides could have accounted for ∼15–20% of total glycerol added to the bloodstream as it passes through abdominal adipose tissue. In addition, systemic glycerol R_a could miss 10–20% of whole-body adipose tissue lipolysis because of the inability to detect lipolysis of intraperitoneal fat, when glycerol released into the portal vein is cleared by the liver (49). Therefore, measuring systemic glycerol R_a by infusing a labeled glycerol tracer probably provides a reasonable estimate of whole-body adipose tissue lipolytic activity under postabsorptive conditions that simultaneously overestimates (glycerol released by lipolysis of plasma triglycerides) and underestimates (glycerol released by lipolysis of intraperitoneal triglycerides) selected areas of regional adipose tissue glycerol release.

In summary, the data from this study show that glycerol released during lipolysis of adipose tissue triglycerides moves along a concentration gradient from interstitial space, to local veins, and into the systemic circulation. The calculated capillary diffusion capacity (ps) for glycerol increased with increasing adiposity and was lower than values previously proposed by assessments made in vitro.

This study was supported by National Institutes of Health Grants DK37948, DK56341 (Clinical Nutrition Research Unit), RR00036 (General Clinical Research Center), and RR00594 (Biomedical Mass Spectrometry Resource), The Special Trustees of St. Bartholomew’s and The London School of Medicine, and The Wellcome Trust.

We thank the staff of the Clinical Research Centre for help with the studies and the volunteer subjects for participation.