Glycerol is synthesized and secreted by adipocytes to dispose of excess glucose, via glycerogenesis and increased acyl-glycerol turnover

White adipose tissue (WAT) produces large amounts of lactate and glycerol from glucose. We used mature epididymal adipocytes to analyse the relative importance of glycolytic versus lipogenic glycerol in adipocytes devoid of external stimuli. Cells were incubated (24/48 h) with 7/14 mM glucose; half of the wells contained 14C-glucose. We analysed glucose label fate, medium metabolites, and the expression of key genes coding for proteins controlling glycerol metabolism. The effects of initial glucose levels were small, but time of incubation increased cell activity and modified its metabolic focus. The massive efflux of lactate was uniform with time and unrelated to glucose concentration; however, glycerol-3P synthesis was higher in the second day of incubation, being largely incorporated into the glycerides-glycerol fraction. Glycerophosphatase expression was not affected by incubation. The stimulation of glycerogenic enzymes’ expression was mirrored in lipases. The result was a shift from medium glycolytic to lipolytic glycerol released as a consequence of increased triacylglycerol turnover, in which most fatty acids were recycled. Production of glycerol seems to be an important primary function of adipocytes, maintained both by glycerogenesis and acyl-glycerol turnover. Production of 3C fragments may also contribute to convert excess glucose into smaller, more readily usable, 3C metabolites.


Results
Isolated adipocytes glycerol release to the incubation medium. Table 1 shows the initial (glucose) and final concentrations of glucose, lactate, glycerol and NEFA in the medium after 24 or 48 h of incubation. Glucose levels steadily decreased and both lactate and glycerol increased during the incubation. The presence of NEFA in the medium also increased dramatically from 24 to 48 h. However, in all cases, NEFA levels were only a fraction (when compared in molar units) of that of glycerol. Figure 1 shows the effect of initial glucose concentration on its uptake by the adipocytes and the efflux of NEFA and glycerol per cell over time. Glucose uptake was lineally dependent on the time of incubation, but independent of medium glucose. NEFA efflux was low during the first 24 h of incubation, markedly increasing when the whole 48 h period was analysed, showing high efflux rates, and no significant effect of glucose concentration.
Medium glycerol was also dependent on the time of incubation, but not on the initial glucose levels, the efflux rates practically doubling glycerol appearance in the medium. The molar ratio of NEFA to glycerol in the medium after incubation was far from the canonical value of 3 (the ratio in TAG) corresponding to pure lipolysis, being in the range of 0.07 to 0.36, the lowest values corresponding to the initial 24 h of incubation. The ratios for the efflux rates showed the same values.
Fate of glucose label. Table 2 shows the distribution of the label, initially present only in glucose, distributed after incubation in the different metabolite fractions. A large proportion of the initial glucose was recovered intact after 24 h or 48 h. These data agree with the fairly uniform rate of glucose uptake by adipocytes, essentially independent of medium glucose concentration. The largest individual label fraction was recovered as lactate. The results obtained with "cold" glucose presented in Table 1 and Fig. 1 are paralleled by the labelled data. The 14 mM glucose groups showed a significantly higher accumulation of label. At 24 h, free glycerol share was highest than that of glycerides-glycerol (especially in the 7 mM glucose group); the differences disappearing at 48 h. The possible complete oxidation of glucose for energy, may be considered improbable, since the label recovered was in the range of 95%.
A significant proportion of label, corresponding to an unidentified fraction (up to 20% of that of used glucose) was found in the medium, especially after 48 h of incubation; the values at 24 h were much lower. We could not identify the nature of this important fraction, not previously detected 36 . We are certain that it is not an acid (i.e. pyruvic, which is retained into the "lactate" fraction), and were neither glycerol (already measured) nor CO 2 , since the data were not related to the estimated production of CO 2 . The results were, then, incompatible with mitochondrial oxidation of Acetyl-CoA. NEFA also were an improbable option, because they would be retained by the columns, more because of lipophilic binding than because of its limited acidity at the pH of extraction. In addition, label in fatty acids, despite its considerable increase in concentration in the medium had a very low specific activity that could not justify not even a small fraction of the label in this important new fraction. Alanine could be a fair candidate, but the source of N was limited.
The values for CO 2 were calculated from the minimal amount needed to incorporate the radioactivity found in the labelled fatty acids fraction. For that reason, we counted this label together with that found in fatty acids and considered the sum as the fraction of label that went through the lipogenic pathway (i.e. 18-36% of total label), values comparatively lower than those retained as 3C units, most of which was returned to the medium (46-66%), probably in the range of 70% if the unknown medium factor is definitively confirmed to be alanine. Figure 2 shows the changes in specific radioactivity experienced by the label fractions isolated and identified after 24 h or 48 h of adipocyte incubation in the presence of glucose. To facilitate the comparisons the data have been plotted on a logarithmic scale, with a value of 1 given to the specific radioactivity of the labelled glucose added to the medium. The specific radioactivity of glucose showed no changes with time, remaining all the time at values not different from 1 (i.e. 10 0 ). The values for lactate tended to show a limited decline with the time of incubation (only for glucose 14 mM), but the change was not statistically significant. Neither were the changes experienced by the medium glycerol, despite a clear trend to decrease with time and higher glucose concentration. The variability of the measurements was considerable, especially for the data obtained with 14 mM glucose. The effect of incubation time was, however, statistically significant for the 7 mM glucose group (P = 0.0479, Student's t test). The decrease in free glycerol specific radioactivity contrasted with the marked, significant increase in glyceride-glycerol data (in any case more than one order of magnitude lower than glycerol). The increase in TAG-incorporated glycerol attests to a marked flow of newly synthesized glycerol into the cell lipid stores, whereas, the decrease in free glycerol shows that only part of this free glycerol can be a direct product of glycerogenesis, the rest being produced via lipolysis of the TAG, which glycerol had a much lower specific radioactivity: it was free of label when the incubation started.  Table 1. Medium levels of glucose, metabolites and cell counts. The data are presented as mean ± sem of eight different two-rat pools (i.e. labelled + parallel). *Estimated values (cell counts were obtained from combined "parallel" well samples). The adipocyte % of lipid (990 g/L) was measured using tissue pooled samples as previously described 35 . The levels of cell TAG were calculated from their lipid content; a standard molecular weight of 884 (i.e. trioleoyl-glycerol) has been used for the calculations. Statistical significance of the differences between groups (2-way-ANOVA). P T represents the effect of time of incubation and P G the effects of initial glucose in the medium.   Table 2. Percent distribution of label from metabolized initial glucose in the main metabolite fractions of rat epididymal adipocytes in primary culture. Values calculated using only the "labelled" well data. Total glycerol corresponds to the sum of the label in medium free glycerol plus acyl-glycerides-glycerol. Statistical significance of the differences between groups (2-way ANOVA). Total "lipogenic" label includes that of cells esterified fatty acids and the calculated minimum CO 2 needed for their synthesis as explained in the text. The effect of "incubation time" was significant for glucose metabolized (P T < 0.0001), glycogen (P T = 0.0087) and for other medium labelled compounds (P T < 0.0001), whereas the significance of "initial glucose concentration" affected only the glucose metabolized (P G < 0.0001). No significant interactions were observed except for metabolized glucose (P i = 0.0021).

Figure 2.
Carbon specific radioactivity of the main label fractions obtained after incubation of epididymal adipocytes in a primary culture in the presence of 14 C-glucose. The data are presented as mean ± sem of four different rats, and are presented in a log scale to show the wide differences between fractions. C-specific radioactivity correspond to the quotient of label found in the fraction divided by the molar concentration and the number of carbons the compound contains. In this case, all data have been referred to initial glucose C-specific radioactivity, to which a value of 1 (i.e. 10 0 ) was given. Blue dots and lines: incubation in 7 mM glucose; green dots and lines: 14 mM glucose. The statistical significance data and conventions are the same as in Figure 1. Figure 1d shows an approximation to the glycerogenic and lipolytic origin of the free glycerol in the medium calculated from the mean values of Fig. 2. In the 7 mM glucose group, at 24 h, practically all free glycerol had been synthesized from glucose, but at 48 h, practically no additional glycolytic glycerol was produced, and the surge in medium glycerol was fuelled by lipolysis. At 14 mM glucose, the pattern was the same, but at 24 h of incubation, lipolytic glycerol was about twice that of direct glycerogenesis. It is remarkable that the pattern of glycerol efflux, shown in Fig. 1e, was the same, irrespective of the availability of glucose in the medium. The specific radioactivity of glycerides-glycerol was 1-2 orders of magnitude higher than that of fatty acids. In addition, the changes described for glycerides-glycerol with incubation time were not observed in the esterified fatty acids fraction. The data agree with lipogenesis being arrested after 24 h in contrast with massive incorporation of labelled glycerol into TAG.
In any case, lipolysis diluted the specific radioactivity of glucose-derived glycerol, but increased its efflux. In spite of lipolysis being the source of part of glycerol, this was not translated into the secretion of NEFA in the high proportions expected. Pure lipolysis produces 3 moles fatty acids per mole of glycerol, but the results were just the reverse, about 3 moles of glycerol per mole of NEFA. Since only part of free glycerol was of lipolytic origin, this ratio may be lower (2-2.5 times more glycerol than NEFA), but in any case was far from that expected for a complete lipolysis. Since comparisons of specific radioactivity were done in terms of C content, not moles, the relationship was drawn even further away. A mean fatty acid has 6-fold more C than glycerol: i.e. 18 to 3. Consequently, the label per C in TAG could not correspond to lipogenic activity matched to the large amounts of newly incorporated glycerol to glycerides, which prompts us to speculate that glycerol turnover in the cell TAG droplet should be much faster than expected. The incorporation of fatty acids newly synthesized from glucose would represent only a fraction of those used to re-synthesize TAG, since most of them were simply recycled, in contrast with the one-way-out of the lipolysis-generated glycerol. Figure 3 shows the levels of expression of transporters, enzymes and other proteins related to the metabolism of glycerol/glycerol-3P in adipocytes, already presented in Table 3, and depicted in the metabolic map of Fig. 4. The data are expressed as the approximate number of copies of the corresponding mRNA per cell, and are presented in a logarithmic scale to allow for comparison of the levels of expression in addition to the trends of change elicited by glucose concentration and incubation time.

Analysis of gene expression of glycerol metabolism-related proteins.
The gene for glycerol-3P dehydrogenase, Gpd1, presents a sizeable number of copies per cell, which increased by one order of magnitude in one day (48 h vs. 24 h) of incubation; no effect of glucose concentration was observed. The glycerol phosphatase gene (Pgp), also showed a high basal number of copies, and a moderate (albeit significant) increase with incubation time. Again glucose availability did not affect the expression of the enzyme. Glycerol kinase gene (Gk) presented a low number of copies compared with Gpd1 and Pgp, but also increased its expression with incubation time and was not affected by glucose. The incorporation of glycerol-3P to form acylglycerols by glycerol-P acyl-transferase (Gpam) showed a similar pattern to Gpd1 and Gk, with a number of copies per cell similar to the latter. Again no effect of glucose concentration in the medium was observed, but incubation time increased the expression of the gene, theoretically facilitating the synthesis of acyl-glycerols if sufficient substrate was available.
The main glycerol transporter, aquaporin 7 (Aqp7), presented a high number of copies per cell, especially when the cells were exposed to 14 mM glucose, and was affected differently by incubation (decreasing under high glucose and increasing when it was low). The 3P-glycerate dehydrogenase gene (Phgdh) is not considered a control node in the glycolytic pathway, but its function is critical to allow the flow of C from trioses-P to pyruvate. In this case, there was a clear interaction between incubation time and glucose availability, observable only at 48 h, in which higher glucose resulted in less marked increases in gene expression. The malic enzyme gene (Me1) related to lipogenesis as NADPH provider, repeated the pattern of increase in expression with time of incubation with nil effect of glucose concentration; its pattern and level of expression being similar to that of Pgp.
The three main WAT lipase genes showed the same pattern than Gpd1, despite acting in the opposite direction of glycerogenesis and glycerol incorporation into TAG. Hormone sensitive lipase (Hsl), and adipose TAG-lipase (Atgl), but also lipoprotein lipase (Lpl) showed a large (highest for Lpl) initial number of copies that increased considerably in the second day of incubation, with nil effects of glucose levels. The gene (Cpt1), coding for carnitine-palmitoyl transferase, which allows the entry of acyl-CoA into the mitochondria, usually for its oxidation, also showed this increasing pattern with time, but glucose also increased its expression (at 7 mM vs. 14 mM), with significant interaction between time and glucose. In any case, the number of copies was very low, which hints at both a complex regulation and limited activity under the metabolic circumstances studied. CD36, one of the main fatty acid membrane transporters, repeated the same pattern of lipases, with similar high number of copies of its mRNA per cell, which may favour uptake rather than release of NEFA.
The expression of pyruvate dehydrogenase kinase 4 gene (Pdk4) was especially interesting. Its small number of copies may be explained by its regulative function on an enzyme, pyruvate dehydrogenase. The increase of almost one order of magnitude from 24 h to 48 h suggests a similarly powerful effect on the synthesis of acetyl-CoA from pyruvate, effectively blocking mitochondrial lipogenesis, and thus the complete oxidative utilization of glucose. This effect was also dependent on glucose concentration. Probably, the changes in Pdk4 were not related to mitochondrial alterations, since the expression of cytochrome C (Cycs), a key mitochondrial marker, remained unaltered by glucose and/or time of incubation.

Discussion
Using primary cultures of mature epidydimal adipocytes, we have found that under conditions of maintained glucose availability (even markedly hyperglycaemic), the cells convert a sizeable part of glucose to 3C metabolites such as lactate and glycerol. The use of 14 C-labelled glucose as precursor has proven the mostly glycolytic origin of the free glycerol released to the medium. The rate of glycerogenesis was largely in excess of the cell needs of sn-glycerol-3P for the synthesis of acyl-glycerols, since the rate of lipogenesis from glucose did not match the larger flow of labelled glucose-C towards the synthesis of glycerol.
A key question for the credibility of this investigation is the validity of the methodology used, thus we invested considerable time and resources to establish its effectiveness and limits. A classical method 37 for adipocyte isolation was adapted, checked and complemented to obtain a basic system of incubation with relevant inclusion of quantitative factors and control of viability 35 . The surge in selective expression of enzymes and transporters, and increased metabolite handling was, in itself, an additional (albeit indirect) proof of the metabolic viability of the cells during a two-day incubation. The use of labelled glucose, a critical point to discern the origin of glycerol and the fate of glucose, was the subject of another previous specific methodology paper 36 .
The main novelty of the present study lies on the combined use of the methodology primarily developed for this investigation, and the combination of different quantitative data obtained from the same source (levels of metabolites, cell counting and size estimation, label distribution and gene expression analyses). The methodological complexity and the large number of data obtained from the same sources, at the same time and conditions, facilitates comparisons, but do not preclude the existence of problems. We believe that the main weaknesses of the present study are: (1) Constrictions affecting the number of samples studied, pooled in pairs. (2) The finding of a large fraction of unidentified labelled compound(s) released to the medium in parallel to the lipolytic surge; we have indications, that this fraction contains alanine (unpublished results). (3) Absence of data on NEFA specific radioactivity (too small samples, and low expected fatty acid label). (4) The non-viability of measuring the small amounts of evolved 14 CO 2 in an atmosphere containing already 5% CO 2 , allowing us only to calculate the minimum cost in CO 2 of lipogenesis; in any case, this figure should be low, given the small proportion of label not accounted for. The data are presented as number of copies of the corresponding mRNA per cell, and are mean ± sem of data from four rats. The data were obtained from the "parallel" incubations (i.e. no label). The results are shown in a log scale to facilitate comparisons of copies per cell between genes and groups. Blue dots and lines: initial 7 mM glucose, mauve dots and lines: 14 mM glucose. The statistical significance data and conventions are the same as in Figure 1. The correspondence between gene names and those of the proteins they code are given in the text and can be seen in Table 3.
(5) The often large variability of some of the label fraction measurements, resulting in statistical uncertainty. (6) The need to use "parallel" wells with no label for the measurement of protein expressions. (7) Too many interdependent results showing complex interactions, which forced us to limit the data presented and discussed here.
The results of our study support an active role of WAT in the handling of glucose, probably helping maintain glycaemia. The main findings were: (A) Glucose uptake was higher than the actual cell energy needs, since even in the absence of insulin, the glycolytic production of lactate apparently provided sufficient ATP to sustain the adipocyte under practically anaerobic conditions 33 . This process is characterized by an increased expression of Glut1 (independent of insulin 38 ) and the production of lactate 33,34 , being the main cell energy-sustaining pathway. The regular, and quantitatively significant conversion of glucose to lactate has been linked to the synthesis of ATP, ADP availability being postulated as the main regulatory factor 30 . Lactate efflux proceeds at a steady pace within a wide range of medium glucose levels (7-14 mM in this study), which agrees with the automatism of the anaerobic metabolism of the thin cytoplasm layer stretched between cell membrane and the lipid droplet surface. Thus, lactate production helps sustain the basic energy needs of most of the cell through a fully anaerobic process 33 . It seems that this source of ATP may suffice to sustain the activity of the small amount of "live cytoplasm" of mature adipocytes 35 . Lactate secretion by WAT may be considered, thus, primarily a normal consequence of the need for ATP generation, and not a specific indicator of hypoxia, despite the generalized association of lactate to hypoxia 39 . In this sense, this mechanism to obtain energy may be more comparable to the Warburg effect of cancer cells 40, 41 than to hypoxia.
A marked glycerogenic flow provided glycerol as a final 3C export substrate together with lactate. A direct extrapolation to the in vivo situation in which lactate is released in large amounts from WAT 42 may hint at this tissue helping lower glycaemia, in fact breaking up 6C to 3C molecules. Glucose was substituted in large proportions by less-regulated 3C substrates, used elsewhere for energy or, eventually, for splanchnic lipogenesis or gluconeogenesis. The existence of a phosphatase directly hydrolysing glycerol-3P was previously postulated by us 30 ; recently, a new glycerol-3P phosphatase has been described in liver, which is also present in WAT 28 . We have found that the corresponding gene was robustly expressed in isolated adipocytes, which agrees with the high glycerophosphatase activity of WAT 29 . The expression of the enzyme (Pgp) increased with time but not with the concentration of glucose; and was neither correlated with the rates of synthesis and efflux of glycerol.  The expression of Pgp seems to be affected by diet and by the lipolytic/lipogenic orientation of the specific adipose tissue analysed 28 . The small, but significant, rise in expression observed here hints at a modulated response. Perhaps the phosphatase activity is more dependent on hormonal control than on bulk substrate. Glycerol phosphatase provides the most direct (and specific) known mechanism to control the availability of glycerol-3P for synthesis of acyl-glycerols, via modulation of the direct hydrolysis of the phosphate ester cosubstrate. The production of free glycerol through this process has been demonstrated in yeasts and plants 43 , where it is catalysed by an enzyme which structure is closely related to that encoded by Pgp 28 . Further study of modifying factors (i.e. exposure to hormones or marked inflow of fatty acids) other than simple glucose availability is needed to check/ understand the role of glycerol phosphatase in the control of the glycerol-3P node. The relatively high number of copies found in comparison with those of glycerol kinase and the also high numbers for aquaporin all point to a clear predominance of the phosphatase over the kinase 4 and the effective removal of glycerol from the cell by aquaporin 7 44 .
(B) The synthesis of acyl-glycerols is a highly regulated process 32 which increased with time of incubation, incorporating large amounts of newly formed glycerol (from glucose) into TAG. Using label distribution data, we found that glycerogenesis was more active than lipogenesis in isolated mature adipocytes, at least when cultured with sufficient glucose. Fatty acids synthesis used only about 1/4th of metabolized glucose (half of its carbon being lost as CO 2 ). The massive efflux of glycerol in cultured adipocytes has been attributed to a non-lipolytic origin, in part because it was not accompanied by a parallel secretion of NEFA 30,45 . Glucose was postulated as the source of glycerol released into the medium by adipocytes 46 , and our results with labelled glucose confirm this origin. The sheer size of adipocytes, and the stretched layer configuration of most of its cytoplasm around the huge lipid vacuole, physically hampers the timely intracellular circulation of substrates. The long (peripheral) distances, the limitations of cytoplasmic currents in adipocytes due to simple geometry, and the rates of diffusion limit most metabolic activities. The resulting layered microenvironments are the consequence of almost unsurmountable difficulties for fast and continuous cytoplasm/mitochondrion interactions, such as pyruvate oxidation and lipogenesis. In most of the cell, glucose or fatty acids can be taken up easily from the interstitial space, and the glucose converted anaerobically to lactate, pyruvate or glycerol, with minimal needs of ATP. But the production of acetyl-CoA requires access to mitochondria, sparsely distributed on large adipocytes 47 . This is not the case with small or growing cells, such as the 3T3L1 converted fibroblasts 48 , where mitochondria and multiple fat vacuoles are interspersed in the surrounding cytoplasm. The physical constrictions may help explain why, in mature adipocytes, glycerogenesis and incorporation of exogenous fatty acids prevail over lipogenesis 45 .
(C) The adipocytes are able to redirect the glycolytic flow towards lipogenesis, glycerogenesis or oxidative metabolism according to their size/geometry limitations and exposure to glucose, irrespective of the concentration of the sugar. These changes were spontaneously activated by adipocytes in the absence of external stimuli other than glucose, and/or the products of its catabolism. We postulate that the coordinated changes (and their direction) observed may be part of a fail-safe automatic mechanism established in the adipocyte to maintain metabolic control against an excess of substrates even in the absence of external regulatory signals. In our study, the absence of insulin did not affect the maintained incorporation of glucose by the cell, and neither lipogenesis, which is known to depend on insulin 49 .
In the present experimental setup, lipolysis was activated by exposure to glucose, without other external stimuli. Glycerol-3-P fate shifted, in part, from being essentially hydrolysed yielding glycerol to being incorporated into acyl-glycerols. This process, however, decreased the availability of glycolytic NADH, needed to convert pyruvate to lactate, thus increasing the availability of pyruvate for oxidative decarboxylation to acetyl-CoA. This process was markedly hindered not only by cell geometry, but also by the marked rise of the expression of Pdk4, an inhibitor of pyruvate dehydrogenase. The consequence was a decrease in lipogenesis in spite of the excess pyruvate available. The absence of an increase in TAG-fatty acids label of adipocytes (in comparison to their glycerides-glycerol) is further proof that lipogenesis practically ceased after the first 24 h coinciding with Pdk4 activation.
In the cell TAG stores, the amount of label incorporated as glycerol was of the same order of magnitude than that of fatty acids (similar number of labelled carbons, not molecules). The stoichiometry of production of one glycerol molecule for each pyruvate, and the utilization of the latter for the synthesis of acyl-CoA via acetyl-CoA is suggestive of lipogenesis as some sort of automatic process for disposal of pyruvate. The synthesis of additional acyl-CoA could be more a consequence than a key objective for disposal of glucose carbon. The glycerogenesis process, we postulate, would modify the glycolytic pathway to produce net pyruvate (not lactate) and excess glycerol-3P. This situation may facilitate both lipogenesis and the synthesis of TAG, provided that glucose supply is maintained. This combination of mechanisms has not been described before, but is supported by the results: in mature adipocytes, the existence of (aerobic) lipogenesis, fuelled by (anaerobic) glycolysis (in the absence of insulin), results in active TAG turnover, sustained by glycerogenesis.
(D) The outflow of glycerol does not follow the steady glycolytic rhythm shown by lactate efflux (unpublished results). Over time (in the second day of incubation), lipolytic-origin glycerol largely substituted direct glycerol-3P hydrolysis as main source of medium glycerol. This was the consequence of a marked rise in lipolysis, which was not paralleled by a matching efflux of fatty acids. Medium NEFA levels increased considerably, but in a proportion much lower than that of glycerol, even when only lipolytic glycerol (and not that coming directly from glycerol-3P) was taken into account. We assumed that most fatty acids freed by intracellular lipolysis were recycled. And those eventually produced by lipoprotein lipase from droplets or exosomes were largely incorporated into cell TAG with freshly formed glycerol-3P; this extracellular lipolytic glycerol adding to that released from the cell via aquaporin 7 50 . In sum, glycerogenesis from glucose shifted from massively freeing glycerol (necessarily via phosphatase) to increase its incorporation into TAG which turnover freed even more glycerol.
The contradictory coexistence of increased lipolysis (proven by the decreasing specific radioactivity of glycerol efflux) and increased synthesis of acyl-glycerols (enhanced glycerol label incorporation), plus higher lipogenesis (ultimately from glucose, as shown by the label found in the fatty acids of TAG) can only be explained by an increase in TAG turnover. This may be considered an example of "futile cycles" spendthrift mechanisms postulated to provide ways to eliminate excess energy, such as thermogenesis. Another postulated futile cycle, based on glycerol kinase was found to be activated by thiazolidinediones 51 , but is actual operation, i.e. free glycerol recycling, has been refuted 16 . However, the steady production of glycerol, and the sequentially compensatory action of the phosphatase and TAG turnover paths, suggest that glycerol synthesis from glucose and its release from adipocytes may be an objective in itself, irrespective of the mechanism used. The main and primary consequence of this process was the net release of free glycerol. That is, glucose-derived glycerides-glycerol was freed by lipolysis, but most of the fatty acids were recycled. Probably, the justification of lipogenesis may be, at least under these conditions, only a consequence of enhanced glycerogenesis and the equilibrium of NADH usage in the cytosol of the adipocyte (unpublished results). Perhaps this glycerol plays an important role elsewhere, as has been suggested for heart normal operation 22 . This hypothesis is also supported by the effort/energy expense devoted to its massive production and release by the adipocyte through two different complementary (sequential?) pathways (phosphatase and TAG turnover). This is a critical open point that deserves further detailed experimental investigation.
(E) We had postulated previously that adipocytes (or WAT) take up more glucose than needed when confronted with high glucose levels, converting a large proportion of it into 3C fragments, such as lactate 52 , pyruvate 53 , alanine 54 and glycerol 1 . These 3C units may be used as energy substrate elsewhere; largely, by the liver in the gluconeogenic 5 and/or lipogenic pathways 55 . But with this action, WAT also disposes of (or defends from) an excess of glucose that may damage its function by dramatically inducing an inordinate enlargement its TAG stores 33 . This is part of a defence process that includes the limitation of blood flow as part of its protection against excess energy substrates 56 . Since WAT accounts for a sizeable part of body mass, and produces large amounts of lactate, pyruvate, glycerol an (probably) alanine, blood glucose levels should decrease, thus helping lower the inflammation and other damaging (i.e. glycosylation) effects caused by hyperglycaemia. The entry of 3C fragments in most tissues goes unhindered by insulin resistance and the tight control of glucose uptake 57 . This approach provides ready to use energy substrates, which are already partially metabolized in a way comparable to that of 3-hydroxybutyrate vs. NEFA or TAG. These fragments are massively used by liver 5,55 , muscle 58 , heart 59 , brain 60 and other tissues, including the adipose tissues (WAT, BAT) themselves 61 .
The main purpose of all these processes may be summarized in the contribution of WAT to decrease the glycaemic load of the body 62 ; of all the glucose consumed by the adipocytes, about 70% found its way into glycerol, lactate and other metabolites. We included here the portion we suspect corresponds to alanine and that of cell metabolites, largely glycogen, fairly abundant in WAT in relation to live cell volume 63 . In contrast, only about 10% was found as fatty acids. Despite the probable errors and variability in accounting, after discounting the losses and estimated CO 2 production, most of the glucose was simply converted to 3C units. This is indeed a remarkable feat that goes against the general assumption that most of the glucose arriving at the adipocyte is inexorably converted to fat by the cells' lipogenesis-oriented metabolism.
The amount of glucose managed by the adipocytes is considerable, in spite of its small active cytoplasm proportion (in the range of 1% of tissue mass) 35 . The large mass of body WAT reinforces the postulated importance of this tissue in the control of glycaemia.
The uniform proportion of glucose taken up and converted into 3C, irrespective of glucose concentration, points towards an intrinsic automatic mechanism of compensation (and, perhaps, protection). The process could be modulated by the mass of substrate available rather than by external regulatory factors. This may be part of a fail-safe mechanism that takes place under conditions of generalized deregulation. If this hypothesis is finally proven, then WAT would be more of a protagonist of energy triage than the obliged recipient (depot) of excess energy 64 . The signalling role of glycerol has been analysed 65 , and WAT is the choice organ source for its release 66 . However, this line of thought needs a more complex experimental scheme to discuss, or even to allow us to speculate further. In any case, it remains a troubling idea, which may in the end move us to reconsider the unanimous assumption of the pathologic nature of WAT accumulation, as, simply, a partly derailed element of a defence system unable to cope with a disordered availability of substrates. The alternative interpretation of an actual effective defence function is in concordance with the beneficial effect of insulin resistance in starvation becoming the basis of type 2 diabetes under conditions of excess.
In sum, we have found that mature adipocytes in primary cultures synthesize and release lactate and a large proportion of glycerol. The latter is a mechanism that needs some time of exposure to glucose to elicit a massive glycerogenic response, parallel to the synthesis and release of fatty acids, albeit in markedly lower proportions. This is paralleled by matched changes in gene expression. The pattern of change was different from the uniform rates of lactate production, unrelated to the concentration of glucose. The changes in glycerol production were paralleled by deep modifications of the enzymes of glycerogenesis and utilization of glycerol-3P. However, the expression of a robust glycerophosphatase was not modified by glucose availability. The stimulation of glycerogenic enzymes was mirrored by similar increases, with time, in WAT main lipases, and largely substituted glycolytic glycerol by the lipolytic product of TAG turnover. This turnover contributed to a higher efflux of glycerol (and, partially, of NEFA), while recycling most of the fatty acids, in a process far from being energetically efficient when compared with lactate production. Consequently, it is postulated that production of glycerol is an important primary function of adipocytes, supported by glycolysis and TAG turnover. Both lactate and glycerol production are assumed to contribute significantly to convert glucose to 3C units, thus lowering the negative effects of excess glucose.

Methods
Rats, housing, handling and sampling. All animal handling procedures and the experimental setup were in accordance with the animal treatment guidelines set forth by the corresponding European, Spanish and Catalan Authorities. The Committee on Animal Experimentation of the University of Barcelona specifically authorized the procedures used in the present study.
Male Wistar rats (Janvier, Le Genest-Saint Isle, France), 14-week old (N = 16), were used after at least 1-week acclimation period. The rats had free access to food (standard rat chow: #2014, Teklad Diets, Madison WI USA) and water at any time, and were kept in two-rat cages with wood shards as bedding material, at 21.5-22.5 °C, and 50-60% relative humidity; lights were on from 08:00 to 20:00.
The rats were killed, under isoflurane anaesthesia, at the beginning of a light cycle, by exsanguination from the exposed aorta. They were rapidly dissected, excising samples of epididymal WAT. Tissue samples of each pair of rats were coarsely minced and pooled. Thus, eight 2-rat samples were used.
For each experiment, two series of incubations were carried out: (a) Adipocytes incubated in the presence of labelled glucose used to determine the glucose fate and specific radioactivity of metabolites; and a parallel group, (b) incubated in the same conditions except for the label, used for cell counting, to analyse gene expressions, and to obtain additional data on media metabolites.
The incubation medium was supplemented with 14  1.8 kBq of 14 C-glucose. Specific radioactivity was expressed in Bq/µmol-C i.e. per micromole of the substrate divided by the number of C in the molecule, thus allowing a direct comparison of specific radioactivity between molecules of different size 36 .
Each well received 400 µL of the cell suspension. Since 0.1 mL of medium was used for initial measurements, the final incubation volume was 2.0 mL. The cell plates were kept at 37 °C in an incubation chamber, ventilated with air supplemented with 5% CO 2 , which gave a theoretical pO 2 of 20 kPa (i.e. 0.2 mM of dissolved O 2 ). These values were in the range of previous experimental measurements done under the same conditions 33 . The cells were incubated for 24 or 48 h without any further intervention, as previously described 36 . Cell recovery, measurements and processing of labelled cell components. The incubation of adipocytes was stopped by harvesting the cells. The medium was pipetted out, mixed, aliquoted and frozen. The procedure for measuring label distribution in the different fractions of cells and media was developed, tested and quantified previously 36 . Briefly, the cells of wells incubated with labelled glucose were weighed, frozen with liquid nitrogen, transferred to glass tubes and immediately extracted with chilled peroxide-free diethyl ether. The aqueous fraction contained small remnants of medium, but essentially cell metabolites and glycogen. The interphase contained most of the cell proteins. The aqueous (and interface) fraction was used whole to estimate its radioactivity. The organic phase, essentially containing TAG, was dried, weighed, re-dissolved in ethyl ether and saponified using KOH in ethanol. The potassium soaps were extracted and counted. The aqueous phase essentially contained only glycerides-glycerol label; it was also removed and counted 36 . Soap label was that of TAG fatty acids. Total cell label was estimated from the cells suspension, TAG label was the sum of total glyceride-glycerol and fatty-acid soaps counts.
The cells of the "parallel" wells were used to extract their RNA for analysis of gene expression. Total cell volume was also calculated from cell numbers and mean cell size. Since cell lipid proportion was known (as indicated in Table 1), we were able to estimate their TAG content 35 , as a way to check (or correct the values in small size samples) the weight of adipocyte ethyl ether-extracted lipid from labelled cells.

Processing of the incubation media.
We used the media of both "parallel" and label-containing wells to estimate the levels of glucose, lactate, glycerol and non-esterified fatty acids (NEFA). We also applied the protocol for labelled metabolite fractioning previously described 36 .
Lactate (including pyruvate) label was determined using centrifuge microcolumns made up with sieve-filter type centrifugation inserts (Ultrafree-MC, Millipore, Bedford, MA USA) containing 250 mg of just hydrated, spin dried cationic-form Dowex 1 × 2 ion exchange resin (Serva Electrophoresis GmbH, Heidelberg, Germany) as previously described 36 . The retained lactate fraction was eluted with acid and counted.
The medium free of lactate was used in part to convert all glucose to gluconate by incubation with glucose oxidase (type VII from Aspergillus niger, Sigma-Aldrich). Catalase (from bovine liver, Sigma-Aldrich) was added to break up H 2 O 2 and help maintain O 2 availability. The change of non-ionic glucose to gluconate allowed its retention (and acidic elution) using microcolumns as those described for lactate. The label retained was that of the unaltered glucose remaining in the medium after incubation 36,69 .
A second aliquot, of the label-containing medium (already free of lactate) was treated with glycerol kinase (from Escherichia coli, #G6278, Sigma-Aldrich) with ATP in a medium adequate for the complete conversion of glycerol to glycerol-3P. The change in ionization was used to remove the glycerol (as glycerol-3P) from the medium using a microcolumn, eluting it with acid and thus counting the label retained in the glycerol moiety 36,70 .
Combination of "cold" metabolite measurements and their radioactivity allowed us to calculate the fate of the initial glucose label under all conditions tested and to estimate the specific-C radioactivity for all of them.
Carbon dioxide production along the lipogenic process was estimated by the calculation of NADPH needed to synthesize one (~C18) acyl-CoA molecule (equivalent to one fatty acid residue in TAG) and assuming that 1 mole of CO 2 was produced in the pentose-P pathway for each 2 moles of NADPH generated (explained in more detail in Ho-Palma et al. 36 ). The label present in TAG fatty acids allowed us to calculate the amount of glucose needed to be oxidized to CO 2 to provide C and reducing power for that synthesis. Since the ratio was constant, (minimum) label in CO 2 was calculated as a correlate of that found in the cell (soaps fraction) fatty acids.
Gene expression analyses. Total cell RNA was extracted from the harvested cells (from "parallel" wells) using the Tripure reagent (Roche Applied Science, Indianapolis IN USA), and were quantified in a ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington DE USA). RNA samples were reverse transcribed using the MMLV reverse transcriptase (Promega, Madison, WI USA) system and oligo-dT primers.
Real-time PCR (RT-PCR) amplification was carried out using 10 μL amplification mixtures containing Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA USA), 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 0.5 for all runs.
A semi-quantitative approach for the estimation of the concentration of specific gene mRNAs per unit of tissue weight was used 71 . Arbp was used as the charge control gene 72 . We expressed the data as the number of transcript copies per cell, in order to obtain comparable data between the groups, given the uniformity of the samples in that aspect. The genes analysed and a list of primers used are presented in Table 3. Their relationships to the metabolic glycerol node are shown in Fig. 4.