Introduction

White adipose tissue (WAT) main substrate is glucose1,2, which uses in part to synthesize fatty acids for its triacylglycerol (TAG) stores3. WAT also takes up unesterified fatty acids4 or those in lipoproteins, released by gut and liver5,6.WAT produces lactate from glucose7,8, in proportions that may account for up to 30% of all body glucose metabolism, as reviewed by DiGirolamo et al.9. Production of lactate is also increased in primary cultures of adipocytes from obese or diabetic humans, accounting for up to 50–70% of all glucose taken up10, in a process dependent, at least in part, on insulin10,11. Lactate release by WAT has been related to obesity and insulin resistance12, since cells from diabetic humans produce more lactate than those of healthy subjects13. Small adipocytes, during the process of increasing fat storage, produce less lactate from glucose than larger cells, containing more TAG14.

WAT lactate production has been assumed to be part of a Cori cycle15 between WAT –as peripheral organ producing lactate from glucose– and liver. The result of such large production of lactate may be an increase in liver gluconeogenesis and/or hepatic glycogen storage16. However, this explanation could not be quantitatively sustained under conditions of excess glucose availability, since it blocks gluconeogenesis and glycogen storage.

WAT lactate release has been often considered a telltale of hypoxia17, correlated with acidosis18, tissue stress19, mitochondrial function disturbances20, endothelial inflammation21 and altered cytokine release pattern22. The paradigm of lactate release as indication of hypoxia has been challenged by direct analysis of in vivo oxygen levels and consumption by adipose tissue23,24, which is relatively low. WAT blood flow is diminished in obesity25, but the tissue could not be considered hypoxic because direct measurement of pO2 proved that the tissue receives sufficient oxygen to sustain its metabolism given the limited needs of the tissue24. Low blood flow may be partly compensated by lactate release, decreasing blood pH, but freeing more oxygen from oxyhemoglobin to counter hypoxia through the Bohr effect26. The decrease in WAT blood flow may be a defensive mechanism, under conditions of insulin resistance, to limit the entry of blood-carried energy substrates, which the tissue has to take up as end-of-the-line dumping place27.

Adipocyte lactate efflux in the presence of high glucose is levels may be due to:

  1. 1

    a simple defense strategy to survive under hyperglycemia, limiting substrate availability to prevent unwanted hyperthrophia through decreased blood flow at the expense of lower oxygen supply. There is no hipoxia because oxygen needs may be minimized using glucose massively through anaerobic glycolysis. This may also help reduce substrate availability, or

  2. 2

    the recourse to anaerobic glycolysis is a Warburg-like28 effect, such as that observed in fast-growing tumor cells29; the cells forsake oxidative phosphorylation to obtain the ATP they need by using large amounts of glucose in glycolysis to lactate30.

In order to discern between these alternatives to justify high lactate production, we used murine 3T3L1 adipocytes exposed to varying glucose concentrations in the medium, under full oxygen availability and determining how these conditions modulate fat deposition, lactate production and oxygen consumption.

Results

Substrate utilization during 3T3L1 adipocytes differentiation

Figure 1 shows the changes in cell volume and biomass (i.e. the sum of cells and debris produced during harvesting) accumulation in 3T3L1 cell cultures, from fibroblasts at confluence to mature adipocytes. Medium glucose (22 mM when fresh), decreased steadily with differentiation in spent media, in parallel to increasing lactate, which arrived at a plateau in the 25 mM range. These trends were maintained when the data were expressed as mean rates per cell. Proton release/leakage increased during differentiation with rates similar to those of glucose uptake.

Figure 1
figure 1

Changes in cell volume, culture biomass, glucose uptake and lactate release of 3T3L1 cells along the differentiation process from fibroblasts (at confluence) to mature adipocytes.

F = fibroblasts at confluence; D1 = end of the treatment with the first differentiation medium; D2 = end of the treatment with the second differentiation medium; A1 = mature adipocytes on day 6; A2 = mature adipocytes on day 9. The composition of the incubation media used in the transitions from one stage to the other are indicated at the bottom, as well as the duration of the process. NBS = neonatal bovine serum; FBS = fetal bovine serum; Ins = insulin; Dex = dexamethasone; IBMX = 3-isobutyl-1-methylxanthine, a phosphodiesterase inhibitor. The upper panel shows the mean cell volumes (±SD) for cells at the different stages of differentiation and growth in size. Since cell volume includes thousands of individual measurements we used the SD instead of SEM values to give a better idea of cell size variability The second panel shows the data for biomass (mean ± SEM) in the culture wells; purple corresponds to cells and black to total biomass, i.e. cells and other smaller particles (largely debris formed during harvesting). The third panel shows the lactate and glucose concentrations (mean ± SEM) in the culture medium (initially lactate 0 mM, glucose 22 mM). The fourth panel presents the mean rates (in attokatals per cell) for lactate production, glucose uptake and proton release into the medium calculated (see text for information) using mean parameter values.

Full size image

Energy balance and glucose fate during the differentiation of 3T3L1 adipocytes

For the sake of calculations, we assumed that the cell components of a fibroblast and a mature adipocyte were similar and the differences in size between them were largely due to accumulation of TAG.

The mean biomass of a confluent fibroblast culture was 1.27 μL/well and that of mature adipocytes obtained from the same stock and number of fibroblasts was 3.59 μL/well at the end of the 9-day process. The difference in volume, assumed to be mainly fat, translated to about 2.07 mg/well, i.e. 2.3 μmol TAG, synthesized using 36 μmol glucose (for details see Supplemental Methods). On the other hand, the 9-day process of differentiation and growth of adipocytes consumed 196 ± 36 μmol glucose, producing 339 ± 4 μmol lactate. The amount of glucose converted to lactate was about 169 μmol, leaving only 27 μmol glucose for TAG synthesis and other uses, including oxidative energy production. Even assuming a wide margin of error in our assumptions, these values were in the same range than those calculated from biomass changes. These data suggest that practically all available glucose was used for lactate production and lipid synthesis.

In the process analyzed, 45% glucose (of the initial 22 mM) was taken up from the medium; about 80–86% of the incorporated glucose was converted to lactate and 15–18% was used for TAG synthesis. From these data we can assume that ATP was mainly obtained from glycolysis: 339 μmol, 36 μmol were excess ATP produced from TAG synthesis and the equivalent of 16 μmol was lost as H+ leakage (assuming 3 protons per ATP). Thus, aerobic glucose oxidation was necessarily very low, since only the pyruvate dehydrogenase step in the lipogenic process was aerobic and anaerobic pathways practically justified the entire glucose uptake. In consequence, 3T3L1 cells showed a markedly anaerobic metabolism under high glucose, even during differentiation and TAG storage.

Figure 2 presents the medium lactate and glucose concentrations after incubation of mature 3T3L1 cells (from days 6 to 9). Glucose levels were practically zero in the spent media from initial values of 0 to 10 mM, reaching up to 17 mM for the 22–30 mM range. The zero values observed below 10 mM suggest that total removal of glucose under these conditions could have gone beyond these values if more glucose were initially available. On the other hand, lactate levels increased up to 20 mM but then leveled off irrespective of glucose availability. The adjustment of medium concentrations to rates per cell repeated the same pattern. There was a significant release of protons, parallel to glucose uptake.

Figure 2
figure 2

Effects on glucose uptake, lactate efflux, pO2 and pH of the concentration of glucose in the medium of 3T3L1 mature adipocytes.

Left upper panel: medium glucose and lactate (mean ± SEM) after 3 days of incubation of the cells at the given glucose concentration. Blue squares represent the initial glucose concentration. The left bottom panel shows the same data presented as rates (attokatals per cell, mean values only); proton release into the medium has been also included (mean ± SEM).The right panels represent the pO2 and pH vs. initial medium glucose concentration (mean ± SEM). NC correspond to a well incubated under the same conditions (i.e. [Glc] = 22 mM) but containing no cells.

Full size image

Effect of medium glucose concentration on adipocyte uptake of glucose and lactate production

No morphological differences between adipocytes exposed to different glucose concentrations were observed (Supplemental Figure 1) and no fibroblasts were observed in any glucose-concentration group. Biomass and cell size/counts were not significantly different either, but there were slight differences in total cell mass and biomass (Supplemental Figure 2). In our calculations, the changes in biomass were equalized by cell counts and resulted in a net production (in 3 days) of −0.13 mg (at glucose 0 mM) to 0.34 mg (at glucose 22 mM). At concentrations of glucose 10–30 mM, the percentage of glucose uptake used for lipid synthesis was in the range of 11–12%, with smaller percentages for lower initial glucose levels (Supplemental Table 1). Lactate was produced (4.2 μmol) even at 0 mM glucose; it was probably formed from alanine or residual glucose from previous medium change. Glucose uptake (with respect to medium glucose) decreased from a maximal 94% at 3.5 mM to 42% at 30 mM. Conversion of glucose taken up to lactate decreased from 120% at 3.5 mM to 76% at 22 mM. The mean proportion of glucose conversion to lactate was 98% and the leakage of protons was equivalent to about 5% of the ATP obtained from lactate.

Figure 2 also presents the values of pH and pO2 in the medium of 3T3L1 cells exposed to different glucose concentrations. The pO2 of wells with no cells was 20.5 kPa, a figure close to that calculated from incubator atmosphere gas composition (20 kPa). The presence of cells resulted in a slight decrease of pO2, in a range of 18.2 kPa to 18.7 kPa, with no differences between the groups, but all of them being lower than the no-cell control well (p < 0.05, Student's t test). The high pO2 values showed that the cells were not under hypoxic conditions.

The medium pH in the no-cell control well was 7.54 and decreased in those with cells, the lowest value, 7.28 corresponding to the highest initial glucose concentrations (p < 0.001 for initial glucose; one way ANOVA). The acidification was due to proton leakage from the cells, proportional to glucose concentration.

Gene expression during the differentiation of 3T3L1 cells to adipocytes

Since the use of Transwells eliminated the presence of non-differentiated fibroblasts in wells containing only adipocytes31, we were able to refer the gene expression data to cell numbers, since the semiquantitative method used for gene expression analysis allowed this type of comparisons32. The results were presented as mean number of copies of the corresponding mRNA per cell. This is only an approximation, but given the homogeneous origin of the samples, they allow comparisons of the expression data between different groups.

Figure 3 shows the expressions of a number of representative genes of glucose metabolism, lipid synthesis, handling and storage and a few regulatory genes controlling these processes or used as signals to other tissues. All genes studied showed a significant effect of differentiation/growth on their expressions, but followed different patterns. The number of copies for lactate dehydrogenase a (muscle type) and fatty acid transporter protein Ap2 were more than three orders of magnitude higher than those with the smallest number of copies: leptin and carnitine palmitoleoyl-transferase. Hexokinase and pyruvate dehydrogenase kinase 4 showed a similar pattern: increased expression during the development period followed by a drop in the largest adipocytes. Preadipocyte factor 1, as expected, showed a reverse pattern, high on fibroblasts and lower in all adipocytes. The rest of genes studied followed a similar time-related pattern, increasing their expression during the different stages of differentiation from fibroblasts to a maximum in 9-day mature adipocytes.

Figure 3
figure 3

Levels of expression of key genes in 3T3L1 cells during the differentiation process from fibroblasts (at confluence) to mature adipocytes.

F = fibroblasts at confluence; D1 = end of the treatment with the first differentiation medium; D2 = end of the treatment with the second differentiation medium; A1 = mature adipocytes (day 6); A2 = mature adipocytes (day 9). The columns represent the mean (±sem) number of copies of the corresponding mRNA per cell at the different stages of differentiation. The statistical significance of changes along the process (one-way ANOVA) is given within each panel. Grey: carbohydrate catabolism pathways; Orange: 3C to 2C conversion; Red: lipogenesis; Blue: TAG catabolism; Purple: fatty acid transport; Green: regulatory agents.

Full size image

Gene expression modulation during exposure to changing glucose concentration in the medium

In Figure 4, the effects of medium glucose on gene expression are shown. Here, the significance of differences between groups was limited to hexokinase, phophofructokinase (muscle type), lactate dehydrogenase a (muscle type), carnitine palmitoleoyl-transferase, hormone-sensitive lipase, adipose TAG lipase, PPARγ, 11β-hydroxysteroid dehydrogenase type 1 and leptin. Most patterns showed slight changes only, with lowest expression values at 30 mM glucose; the typical pattern, better exemplified by hexokinase showed a slight rise from 0 mM to 3.5 mM glucose followed by a progressive decrease in expression with increasing medium glucose. There was a lack of significant changes in all genes coding for enzymes implicated in lipid synthesis in spite of the actual increase in TAG synthesis (i.e. fat stores grew) in parallel to medium glucose. The expression of the main lactate dehydrogenase (muscle type) isoform decreased in spite of increasing lactate production with high medium glucose.

Figure 4
figure 4

Levels of expression of key genes in mature 3T3L1 adipocytes incubated during three days at different glucose concentrations in the medium.

Points represent the mean (±sem) number of copies of the corresponding mRNA per cell. The statistical significance of the effect of medium glucose concentration (one-way ANOVA) is given within each panel. Grey: carbohydrate catabolism pathways; Orange: 3C to 2C conversion; Red: lipogenesis; Blue: TAG catabolism; Purple: fatty acid transport; Green: regulatory agents.

Full size image

The ratio of transcripts per cell between total lactate dehydrogenase and muscle type enzyme gave a mean of 94.3 ± 0.3% of transcripts for muscle type enzyme, i.e. practically all (ratio in the range of 19/20). Thus, 3T3L1 cell lactate dehydrogenase belongs to the muscle type isozyme.

In vivo analysis of rat tissue lactate and lactate dehydrogenase

Table 1 shows the lactate dehydrogenase activity in the rat tissues tested, as well as the expressions of lactate dehydrogenases and a genes controlling fatty acid synthesis. Liver lactate dehydrogenase activity was one order of magnitude higher per g of protein than that of WAT. However, the differences in expression of the genes for this enzyme were much higher, between 3 and 4 orders of magnitude larger: i.e. WAT lactate dehydrogenase activity was related to a much lower amount of copies of both mRNALdha and mRNALdhb than liver in spite of the latter higher specific activity.

Table 1 Lactate dehydrogenase activity and gene expressions in liver and adipose tissue of adult male Wistar rats
Full size table

WAT locations showed similar expressions of both genes coding for lactate dehydrogenases, which suggests that in adult male rats half of the transcripts were of the muscle subtype and the other half of the heart isoform, whilst the liver was overwhelmingly of the muscle type as has been previously established33. All gene expressions investigated showed that subcutaneous WAT had higher expression levels than epididymalWAT, the differences being, however, not significant, in part because of high individual variation. The expression of Acc1 and Pdk4 were one order of magnitude lower in WAT than those of Ldha and Ldhb, but not than that of Fas, which also showed the highest variability.

Table 2 presents the plasma levels of lactate and glucose in blood plasma and the concentrations of lactate in tissue (per g of tissue and as molality fraction), as well as the molality quotients between tissues and plasma. Glycemia was relatively high for normal rats (they were under isoflurane anesthesia, which increases glycemia, when killed), but was maintained within the range of normalcy for this stock. The levels of tissue lactate in liver (μmol/g) were four-fold higher than the circulating plasma levels, but those of WAT were lower. However, when only water space was taken into account (i.e. discarding the mass of TAG), the molality ratios of tissue lactate to plasma showed that in both WAT sites and liver, tissue lactate concentration was higher than in plasma.

Table 2 Lactate concentrations in plasma, liver and adipose tissue of adult male Wistar rats
Full size table

Discussion

A key finding of this study is that lactate production by 3T3L1 adipocytes is not a consequence of hypoxia. Direct measurement of pO2 showed that cultured cells had higher oxygen availability than that of blood plasma. Along the whole process of differentiation and growth, only a fraction of the widely available medium glucose was converted to TAG, whilst a larger proportion of the available glucose was converted to lactate through a fully anaerobic pathway.

Why was so much glucose metabolized anaerobically (i.e. wasted from the point of view of carbon or energy utilization) in the presence of sufficient oxygen to sustain a more efficient aerobic glucose oxidation? Did the cells show a Warburg effect28,30? It is plausible that a Warburg-like mechanism may suffice to cover the cell energy needs under conditions of rapid growth and abundance of energy substrates, since the ATP yield of glycolysis may be enough for rapid cell buildup, as in tumors29. However, the results we obtained do not support a Warburg-like effect for adipocytes, since the ability to use glucose through glycolysis to lactate as end-product was parallel to cell differentiation and lipid accrual. It may be expected to find a maximal Warburg effect at the earlier stages of differentiation, i.e. fibroblasts, but we observed just the reverse. In addition, we found no cell proliferation (cells' numbers did not increase; adipocytes only grew larger), another factor for active glycolysis in the presence of oxygen30.

Lactate production in the presence of varying proportions of glucose suggests that the adipocyte is probably a “obligatory glycolytic” cell. In fact, mature adipocytes produced lactate even at glucose 0 mM (probably from alanine and/or remnants of previous medium). Glycolysis, apparently, took precedence upon lipid synthesis and storage (as shown by the different accrual of biomass). Medium glucose did not affect the expression of enzymes implied in the synthesis of fat, but modified those affecting its degradation and the incorporation of exogenous fatty acids. The strictest control was apparently directed, in fact to limit the entry of glucose into the cells (i.e. hexokinase34) and to lower the expression of lactic acid dehydrogenases, perhaps to regulate glucose wasting. Nevertheless, the amount of glucose broken up to lactate was sustained at high rates even at the highest glucose concentrations.

Uncoupling of mitochondrial oxidative phosphorylation in 3T3L1 cells increases glycolysis to lactate at the expense of lipogenesis35. The sum of the largely anaerobic processes of synthesis of TAG and glycolysis to lactate accounted for practically all glucose. When the protons released (i.e. not used by the mitochondria for energy) were taken into account as their energy equivalent, the conclusion was that 3T3L1 cells had practically nil needs (and, consequently, consumption) of oxygen under the conditions tested.

The lack of changes in pO2 with increasing medium glucose levels also hints at the lack of relationship between the abundance of the cells' main substrate and oxidative metabolism. The data we present here suggest that 3T3L1 adipocytes are essentially anaerobic even during lipogenesis and TAG storage. However, it must be taken into account that WAT contains other types of fully aerobic cells (stromal) in addition to adipocytes and thus WAT as a whole may be more susceptible to hypoxia than its main constituent, adipocytes alone. Our results using 3T3L1 cells can be safely attributed solely to adipocytes31.

Our in vivo data show that WAT lactate dehydrogenase activity was relatively high in comparison with other enzyme activities of the tissue. Lactate molality ratios vs. plasma were always positive, thus, rat WAT produced lactate under standard conditions, in agreement with the existence of lactate gradients in animal and human models14,16. Circulating lactate is increased in the immediate postprandial state, rising later than glucose and insulin36. The process is quantitatively important, since, in insulin-resistant humans, WAT converts to lactate up to 2/3 of all glucose taken up13. The rapid conversion of glucose to lactate has been also observed in vivo, in human subcutaneous WAT16,37, at rates that exceed the relatively low needs for energy of WAT itself. High liver lactate molality ratios reflect its role as lactate receptor in the Cori cycle, since lactate is a prime substrate for gluconeogenesis15.

Adipocytes, by massively breaking up glucose to lactate (especially when WAT mass is large), lower glycemia, thus contributing to maintain glucose homeostasis. The process is mainly sustained by large adipocytes9, thus we can assume that WAT glucose removal (and overall lactate production) are increased in obesity38, in spite of locally reduced blood flow and access to substrates.

High glucose levels tend to decrease the relative ability of 3T3L1 cells to produce lactate from glucose (lower gene expressions); but, in absolute terms, this was not translated into decreased lactate production. We can speculate that the limit of glucose uptake and lactate production may, rather, depend on the ability to use the ATP generated, since the net ATP consumption of lipogenesis is small (Supplemental Figures 3 and 4) and its needs for adipocyte maintenance are limited.

From the data presented we can postulate an additional role for adipocytes in energy homoeostatic regulation: WAT helps decrease plasma glucose, especially under hyperglycemia and thus indirectly improving insulin resistance39. As indicated above, WAT is the last dumping site for glucose under insulin resistance-induced hyperglycemia27,40. In WAT, excess glucose is used, in part, to build up TAG stores, but most of the glucose, unwanted elsewhere, is converted to lactate, which could be used as substrate by a number of tissues41,42. Most of circulating lactate, however, is taken up by the liver43 and oxidized to pyruvate. Under conditions of excess glucose, in the liver, the gluconeogenic pathway remains inhibited44 and the excess 3C is oxidized to acetyl-CoA via pyruvate dehydrogenase, followed by oxidation in the Krebs cycle (if not already saturated), or its utilization in other pathways. Since ketogenesis is also inhibited by glucose45, practically the only path open is lipogenesis, fully operational for lactate in the liver46. Accumulation of newly formed lipid may hamper liver function (steatosis), but liver TAG may be exported (via lipoproteins) to other tissues, including WAT, which thus receives the “unwanted glucose carbon” as lipoprotein-carried TAG fatty acids, aggravating the WAT problem of TAG storage.

In the long term, the process outlined may help extend the damages caused by excess nutrients, but on the short term may help limit the effects of sustained glucose peaks, facilitate energy partition via distribution of 3C substrates and the centralization (and control) of fat synthesis in the liver rather than in disperse WAT sites. The process we postulate is a bidirectional transposition of the well known glucose-fatty acid (or Randall) cycle47.

In conclusion, cultured adipocytes are largely anaerobic and produce huge amounts of lactate from glucose in normoxic conditions. It is speculated that the massive production of lactate, directly related to glucose availability, is a process of defense of the adipocyte, exporting glucose carbon (and the problem) to the liver as lactate, but it may be also a mechanism of control of hyperglycemia. The in vivo data obtained from male rats agree with this interpretation.

Methods

Effects of differentiation of 3T3L1 cells on gene expression and substrate utilization

A batch of 3T3L1 fibroblasts was allowed to grow and differentiate under standard conditions, as described in the Supplemental Procedures, using Transwell™ inserts (Corning Life Sciences)31. Cells were harvested at 0, 2, 4, 6 and 9 days after the beginning of fibroblast differentiation (i.e., day 0 cultures contained only confluent fibroblasts and days 6–9 only mature adipocytes). Cells were harvested, counted and their total RNA extracted using the GenElute™ (RTN10, RTN70 and RTN350, from Sigma-Aldrich) isolation procedure, following the instructions of the provider. RNA was used for the gene expression analysis. Media were used for the analysis of glucose, lactate and proton leakage (as described in the Supplemental Procedures).

Cultured 3T3L1 cells exposure to different medium glucose concentrations

Mature 3T3L1 adipocytes (i.e., on day 6 after fibroblast confluence as described above) were used for the analysis of the effect of medium glucose concentration. The medium was prepared omitting glucose, which was added to final concentrations of 0, 3.5, 5.5, 10, 22 or 30 mM. These media were used the last three days of incubation, substituting the standard 22 mM glucose medium. Spent media were used for the analysis of pH, glucose and lactate concentrations. The cells were harvested, counted and used for the extraction of total RNA for gene expression analysis.

pO2 in the medium during incubation conditions

During the cells' exposure to different glucose concentrations, on day 9 (i.e., before harvesting), the medium pO2 was measured using 1 mL syringes equilibrated at 37°C, which were filled with medium and resealed. Measurements of pH and pO2 were done within 15 min, using an ABL-5 gas analyzer (Radiometer; Copenhagen Denmark). Control wells contained fresh medium and were incubated with no cells.

Analysis of gene expression in cultured 3T3L1 cells

Total RNA from harvested cells in different experiments was quantified using a ND-100 spectrophotometer (Nanodrop Technologies, Wilmington DE USA). RNA samples were reverse transcribed using the MMLV reverse transcriptase (Promega, Madison, WI USA) and oligo-dT primers.

Real-time PCR amplification was carried out using 10 μL amplification mixtures containing Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA USA), equivalent to 4 ng of reverse-transcribed RNA and 150 nM primers. Reactions were run on an ABI PRISM 7900 HT detection system (Applied Biosystems) using a fluorescent threshold manually set to OD 0.150 for all runs. The primers used for the estimation of murine (3T3L1 cells) gene expression are presented in Supplemental Table 2.

A semiquantitative approach for the estimation of the concentration of specific gene mRNAs per cell or unit of tissue weight was used32. Rpl32 was used as charge control gene. The data were presented as the number of transcript copies per cell, allowing for direct comparisons independently ofthe number of cells in a given well.

In vivo study. Animals and housing conditions

Nine week old male Wistar rats (Harlan Laboratories Models, Sant Feliu de Codines, Spain) were used. The rats (N = 6) were housed in two 3-rat cages, had free access to water and were kept in a controlled environment (lights on from 08:00 to 20:00; 21.5–22.5°C; 50–60% humidity) and were fed standard rat chow (2014, Harlan).

The rats were kept, handled and killed following the procedures specifically approved by the University of Barcelona Animal Welfare and Ethics Committee, in accordance with the Rules set by the European Union and the Governments of Spain and Catalonia.

The animals were killed by exsanguination through the aorta under isoflurane anesthesia; samples of liver and WAT at the inguinal subcutaneous and epididymal fat pads were sampled, frozen in liquid nitrogen and stored at −80°C. Blood plasma was separated from (heparinized) blood and kept frozen at −80°C.

Tissue lactate

Frozen tissue samples were weighed and homogenized in about 20 volumes of chilled water: acetone, to a final proportion of 1:1.248. The samples were centrifuged at 0°C for 20 min at 4000 × g. Supernatants were delipidated (when needed) with finely powdered solid MgO and used for the measurement of lactate (in parallel to plasma glucose and lactate) using the same procedures described (Supplemental Procedures) for cultured cell media.

Tissue and plasma lactate concentrations were expressed in molar units per g of fresh tissue weight, but also, as an estimate, per g of tissue water (molality). Adipose tissues contain about 70–80% of fat, which reduces considerably their water (and lactate) space49. The tissue (and plasma) water (i.e., lactate space) was estimated as shown in Supplemental Table 3. Calculation of theoretical tissue lactate concentration in these volumes of water was used to estimate the mean tissue vs. plasma lactate molality concentration ratios.

Measurement of lactate dehydrogenase activity

Frozen samples of liver were homogenized using a tissue disruptor (IKA-T10 basic Ultra-Turrax, IKA, Stauffen Germany) in 10 volumes of chilled Krebs-Ringer bicarbonate solution, pH 7.8 containing 5 mM dithiothreitol, 0.5% bovine serum albumin, 1% dextran (MW 200,000), 0.1% Triton X-100 and 1 mM EDTA50. Adipose tissue samples were homogenized in the same way as liver, but the homogenates were left standing for 5 min at 4°C. Sediments and floating fat-cakes were discarded and only the intermediate layer was pipetted for lactate dehydrogenase activity. Homogenate protein was also estimated51, using homogenization medium for blanks.

Total lactate dehydrogenase (EC 1.1.1.27) activity was measured in liver and WAT homogenates, using a standard kinetic NADH-oxidation UV method52 to determine Vi values. Activity was expressed as nkat/g of wet tissue and μkat/g protein.

Analysis of gene expression in rat liver and WAT

Tissue total RNA was extracted from frozen tissue samples (about 30 mg) using the GenElute™ procedure. Gene expression was estimated as described for cultured cells, using a different set of primers; the list is presented in Supplemental Table 2. The semiquantitative approach described above was used for the estimation of the number of specific mRNA copies for each gene per unit of tissue weight and per g of protein. The genes studied coded for lactic acid dehydrogenases (muscle and heart), two key enzymes of fatty acid synthesis: acetyl-CoA carboxylase and fatty acid synthetase; and pyruvate kinase 4, which controls the use of pyruvate for oxidation to Acetyl-CoA. The Ppia gene was used as control of charge.

Statistical analyses

Comparisons between groups were done using one-way ANOVA analyses and the Bonferroni post-hoc test with the Prism 5 (GraphPad Software, San Diego CA USA) graphics/statistics package. Analysis of correlations and curve fitting were done using the same program. The Student's t test was also used for comparison between isolated data groups.