Mass recovery following caloric restriction reverses lipolysis and proteolysis, but not gluconeogenesis, in insulin resistant OLETF rats

Caloric restriction (CR) is one of the most important behavioral interventions to reduce excessive abdominal adiposity, which is a risk factor for the development of insulin resistance. Previous metabolomics studies have characterized substrate metabolism during healthy conditions; however, the effects of CR and subsequent mass recovery on shifts in substrate metabolism during insulin resistance (IR) have not been widely investigated. To assess the effects of acute CR and the subsequent mass recovery on shifts in substrate metabolism, a cohort of 15-week old Long Evans Tokushima Otsuka (LETO) and Otsuka Long Evans Tokushima Fatty (OLETF) rats were calorie restricted (CR: 50% × 10 days) with or without partial body mass recovery (PR; 73% x 7 days), along with their respective ad libitum controls. End-of-study plasma samples were analyzed for primary carbon metabolites by gas chromatography (GC) time-of-flight (TOF) mass spectrometry (MS) data acquisition. Data analysis included PCA, Pearson correlation vs previously reported variables (adipose and body masses, and insulin resistance index, IRI), and metabolomics maps (MetaMapp) generated for the most significant group comparisons. All treatments elicited a significant group differentiation in at least one principal component. CR improved TCA cycle in OLETF, and increased lipolysis and proteolysis. These changes were reversed after PR except for gluconeogenesis. Plasma lipid concentrations were inversely correlated to IRI in LETO, but not OLETF. These shifts in substrate metabolism suggest that the CR-induced decreases in adipose may not be sufficient to more permanently alter substrate metabolism to improve IR status during metabolic syndrome.

Introduction libitum with standard laboratory rat chow (MF; Oriental Yeast Corp., Tokyo, Japan) for 4 weeks. At 15 weeks of age, rats were separated in two ad libitum food control groups (n = 8/ group/strain), two CR groups (n = 7/group/strain), and two partial recovery (PR) groups (n = 7/group/strain). Both CR and PR groups were subjected to 50% CR (compared to ad libitum control) for 10 days. CR groups were subjected to an oral glucose tolerance test (oGTT) as previously reported [18] to calculate insulin resistance index (IRI) [13] and dissected three days later along with their respective ad libitum controls (CR Ctrl, n = 8/group/strain). Meanwhile, PR groups were fed ad libitum for 7 days, achieving partial body mass recovery (73% recovery of mass loss) before an oGTT and dissection paired with their PR control groups (n = 7/group/strain). Both CR and PR control groups were fed ad libitum for the entire study. However, they were considered independent groups to avoid the confounding factor of age as they were dissected 1 week apart; however, no significant differences were detected between CR and PR control groups despite the slight difference in age suggesting that age was not a confounding factor here.
All animals were maintained in groups of two animals per cage at the start of the study to minimize stress [19] and one per cage during the CR phase to avoid feeding competition. Animals were maintained in a specific pathogen-free facility under controlled temperature (23˚C) and humidity (55%) with a 12-h light, 12-h dark cycle. All animals were given free access to water for the entire study. None of the animals displayed signs of illness before their respective endpoints.

Blood sample and tissue collection
Details of blood sample and tissue collections have been previously reported [18]. Briefly, animals were anesthetized after an overnight fast with a 100 mg/kg i.p. pentobarbital injection. Arterial blood was collected via the abdominal aorta into chilled vials containing a cocktail of 50 mmol/L EDTA, 5000 KIU aprotinin, and 0.1 mmol sitagliptin phosphate (DPP4 inhibitor). Following collection of blood samples, tissues were systemically perfused with PBS. Retroperitoneal and epididymal fat depots were dissected, weighed, and collected for other analyses. Blood samples were centrifuged (3,000 g, 15min at 4˚C), and the plasma was transferred to cryo-vials and immediately stored at −80˚C. Aliquots of plasma (n = 5 per group/strain) were analyzed for primary carbon metabolites by gas chromatography (GC) time-of-flight (TOF) mass spectrometry (MS) data acquisition and processing at the West Coast Metabolomics Center as previously described [20], generating a dataset of 143 consistently identified metabolites. For metabolomics analyses, the 5 samples closest to mean for IRI (previously reported) for each group were included a priori to reduce variability without compromising statistical power (power = 0.81). The metabolomic analyses were performed in a blinded, naïve manner to avoid any potential bias.

Data analysis
Required sample size and actual statistical power was calculated by G � Power 3.1.9.7.
[21]. An n = 7 per group was determined in order to achieve a statistical power of 0.92 with a one-tail α error probability of 0.05 and a Cohen's d of 1.74 (calculated with the mean ± SD IRI of ad libitum fed LETO (8.3 ± 1.2) vs OLETF (19.7 ± 6.5)).
Data normalization and transformation. Data were reported as quantitative ion peak heights and were normalized by the sum peak height of all structurally annotated compounds (mTIC normalization) [22]. Transformations were done using Johnson's family of transformation, depending on data normality, as described in other studies [23]. Student's t-test was performed in 10 previously selected pairwise comparisons ( Metabolic mapping. Fold changes were calculated as mean condition B/mean condition A, after mTIC normalization. The pair-wise comparison of conditions (A-J) are described in Table 1 and referenced throughout the rest of this text. Metabolic maps were plotted per comparison only for metabolites with p<0.05 after Student's t-test using MetaMapp [25] and visualized in Cytoscape 3.7.1 with organic layout. Nodes with significant difference after FDR correction (q<0.2) were highlighted for fold-change and direction of change. A threshold of 0.2 was established for the FDR correction to provide enough statistical power to detect Correlations with body mass, insulin resistance index, and adipose depots. Simple Pearson correlations were calculated for parameters previously reported [18] vs metabolomics data reported for the first time in this study. Johnson transformed body mass, IRI, and retroperitoneal and epididymal adipose masses were treated as independent variables and Johnson transformed metabolites were treated as dependent variables. The independent variables were chosen a priori due to their biological relevance and significant difference across groups, while metabolites involved in carbohydrate, lipids, and amino acid metabolism were included in the correlation analysis. Correlations were considered strong at r<-0.8 or >0. 8   above the threshold were used further for PCA for correlation analyses. The scripts showing the metabolites considered for each PCA analysis can be found in the online repository. The rationale behind using PCA after PLS-DA was to avoid the increased group separation due to artifacts inherent to PLS-DA (found especially in datasets with low sample-to-feature ratios [27]), and instead visually reflect the variability that distinguishes the groups [28]. PCA analysis, although unsupervised, is improved by removing the variables that contribute the less to the separation between groups, which are found by PLS-DA. For each PCA, mean of scores for PC1 and PC2, and standard deviations were calculated, and a t-test performed for both PC for significant difference (P<0.05) ( Table 2). This was performed to demonstrate the separation of the two groups per comparison achieved by PCA in at least one PC. Therefore, the absolute values of the mean PC scores should only be considered as a measure of separation between groups at a specific PC. As the axes for PC1 and PC2 are drawn equidistant to the centroid of both groups in each pairwise comparison, mean PC scores reported in Table 2 are symmetrical at each comparison. For each comparison, the metabolites with the five highest positive and 5 highest negative loadings are shown in Table 3.

Ad libitum fed OLETF rats had lower glycolysis and TCA variables, and higher PPP compared to LETO
OLETF rats had lower basal levels of the glycolysis intermediates, 3-phosphoglycerate (10.9fold, q<0.001), glyceric acid (2.4-fold, q<0.001), and glycerol-alpha-phosphate (2.5-fold, q = 0.001). Moreover, the basal concentrations of citric acid and isocitric acid, the first Table 2. Mean PC score ± SD for PC1 and PC2 for each pairwise group comparison and T-test P value for each comparison. The mean absolute value of the PC score represents the distance from the group centroid to the PC axis, while the sign denotes direction from the axis.  intermediates of the TCA, were reduced in OLETF compared to LETO (1.9-fold, q = 0.032 and 2.0-fold, q = 0.061 respectively). Conversely, basal glucose was higher in OLETF (1.3-fold, q = 0.020), and ribitol and ribose concentrations were increased in OLETF compared to LETO (2.2-fold, q = 0.009 and 1.4-fold, q<0.001, respectively) (Fig 2A). These data suggest that during normo-caloric conditions, PPP may contribute significantly to energy production during an insulin resistant condition.

CR induced higher lipolysis intermediates in OLETF compared to LETO
Compared to LETO after CR, OLETF had higher plasma concentrations of the long-chain fatty acids (LCFA) lignoceric acid (3.1-fold, q = 0.087), myristic acid (1.8-fold, q = 0.125), and palmitoleic acid (2.6-fold, q = 0.038), and the medium-chain saturated fatty acid (MCFA) caprylic acid (2.9-fold, q = 0.103), which could reflect a higher rate of lipolysis in OLETF. Moreover, there is an even more pronounced difference between strains in uracil and uric acid concentration (5.3-fold, q = 0.041 and 4.9-fold, q = 0.021 higher, respectively) (Fig 2B), which suggest either an increase in lean tissue catabolism and/or a deficit in kidney function in the insulin resistant strain.

Amino acid metabolism was more closely correlated with visceral adiposity than total body mass in OLETF after PR
We previously showed that visceral adipose (retroperitoneal and epididymal) depots were reduced with CR, but depots did not recover after PR in OLETF suggesting that lean tissue accounted for most of the mass recovery. Moreover, total amino acids increased exclusively in

PLS-DA with subsequent PCA successfully separated each pair of groups at PC1 except for LETO CR vs OLETF CR
PLS-DA was performed for each significant pairwise comparison before performing pairwise PCA analyses (Fig 4). Each pairwise comparison had significantly different (P < 0.05) PC1 score means, while none of the PC2 score means were significantly different in any of the comparisons. The only exception was comparison C (LETO CR vs OLETF CR) where PC1 scores were not significantly different ( Table 2). Moreover, this comparison had the closest PC1 score means compared to the rest. This could be explained by the relatively higher variability between samples in the OLETF CR group (SD of 5.80 for PC1). In contrast, LETO CR had relatively much smaller variability, (SD of 0.76 for PC1). Based on these observations, and the fact that PC1 accounted for at least 40% of the variation in all comparisons, further PCA analyses focused on PC1.

Carbohydrate metabolism contributed to the distinction between LETO and OLETF ad libitum controls
The monosaccharides, glucose and ribitol, were among the top 5 metabolites that contributed to differentiation between OLETF and LETO (strain effect) on ad libitum diet on PC1 at 16 weeks of age ( Table 3, Comparison A), while hexitol and ribose were added to the list at 17 weeks of age (Table 3, Comparison B). Conversely, the metabolites that contributed to the differentiation between the LETO and OLETF CR control groups were involved in lipid, amino acid, and secondary metabolism, with the exception of 3-phosphoglycerate, which is involved in glycolysis. (Table 3, Comparison A). These changes could imply a higher reliance in glucose metabolism from the OLETF rats or an impairment in lipid or amino acid metabolism during the insulin resistant state.

Amino acid metabolism contributed to the distinction between LETO and OLETF after CR, but not after PR
From the top five absolute loadings, 2 metabolites involved in amino acid metabolism (putrescine and pantothenic acid), and beta-alanine, involved in nucleotide metabolism, contributed to the distinction of OLETF from LETO after CR, while the amino acids, tryptophan, lysine, and oxoproline, as well as arachidonic acid (involved in lipid metabolism), contributed to distinguish LETO from OLETF after CR ( Table 3, Comparison C). This suggests that OLETF rats may rely on protein catabolism in order to synthetize coenzyme A, in contrast to LETO rats that use lipids primarily as a source of energy when carbohydrates are depleted.

PLOS ONE
Conversely, after PR, the metabolites that contributed the most to the distinction between the strains were associated with either lipid metabolism (glycerol-alpha-phosphate for LETO and cholesterol for OLETF) or carbohydrate metabolism (3-phosphoglycerate for LETO, and ribitol and ribose for OLETF) ( Table 3, Comparison D).

Discussion
The previous results show that CR elicited more profound metabolic changes in the insulin resistant state compared to the healthy state. Several of the changes observed in glucose metabolism were expected based on previous findings. However, it was rather surprising that these metabolic shifts were not sufficient to spare the utilization of proteins for fuel in the insulin resistant rats, and even more, that mass recovery did not reverted gluconeogenesis to its basal levels, as we will further discuss.

Carbohydrate metabolism prevailed with an improvement in the TCA cycle, without affecting the oxidative phase of the PPP, after a week of CR in OLETF
A reduction in caloric intake is usually associated with a shift toward increased lipid metabolism because of the reduction in glucose [29][30][31]. Thus, after 7 days of 50% CR in OLETF rats, we expected to observe a shift toward a greater reliance on lipid metabolism [32]. However, the increases in 3-phosphoglycerate, a by-product of glycolysis, several TCA intermediates, and two intermediates from the non-oxidative phase of the PPP, independent of any detectable changes in oxidative phase intermediates suggest that glucose metabolism is increased with CR during insulin resistance. This highlights the clear benefit of CR to improving glucose metabolism and ameliorating the hyperglycemia during insulin resistance. The increase in plasma ribose likely indicates an increase in PPP concomitant with a decrease on nucleotide synthesis in peripheral tissue [33]. The increase in TCA intermediates suggests that CR even during an insulin resistant state may improve the strain-associated impairment in the TCA cycle. This impairment has been previously described in obese, diabetic rats secondary to hyperphagia and an impairment in the leptin receptor [3], and in T2D patients compared to healthy subjects [34]. The improvement in IRI in OLETF with CR [18] substantiates this data suggesting that CR improves glucose metabolism via enhanced TCA cycle activity. However, in the OLETF the impairment in basal glucose metabolism is likely induced by impaired glycolysis because basal levels of 3-phosphoglycerate were lower in OLETF compared to LETO.
Partial mass recovery reverted lipolysis, rather than gluconeogenesis, to basal levels The decrease in glycerol after PR compared to CR suggests that either lipolysis and/or gluconeogenesis is decreased after mass recovery [35], and demonstrates that seven days of partial recovery were sufficient to reverse the limited reliance in lipids in OLETF during an insulin resistant state. This result is consistent with our previous finding where triglyceride levels returned to basal levels following PR [18]. Furthermore, the decrease in the glucogenic amino acids, alanine, cysteine, glycine, and aspartic acid after refeeding (PR), is consistent with the metabolomics data from humans following starvation and refeeding [36]. However, we previously showed strain-dependent changes in the expressions of glucogenic enzymes, although we did not detect consistent changes in these enzymes with CR and PR in OLETF [18] suggesting that the rate of lipolysis, but not gluconeogenesis, is reverted to basal levels in OLETF after PR, Furthermore, comparing our previous measures of hepatic glucogenic enzyme expressions with the current metabolomics data suggests that the lack of detectable changes in protein expression do not accurately reflect the dynamic changes in gluconeogenesis, which puts more of an onus on metabolomics analyses such as these to better complement the molecular data. Additionally, these shifts in lipolysis and gluconeogenesis following PR may highlight the potential detriment of mass recovery, even if only partial as demonstrated here, after CR in an insulin resistant state, as the amelioration of the hyperglycemia with CR may be re-established from the combination of de novo glucose synthesis in the liver [37] and impaired glucose uptake from peripheral tissues [38,39]. If so, this emphasizes the importance of maintaining the reduced body mass after CR as even partial recovery of body mass independent of adiposity has the potential to induce detrimental shifts in glucose metabolism that revert to an insulin resistant, hyperglycemic condition.

Plasma lipids are better correlated to IRI in a non-insulin resistant state than glucose metabolism intermediates
The lack of an inverse correlation between IR and circulating fatty acids (i.e. products of lipolysis) in an insulin resistant state in the present study may be explained by previous findings where insulin was unable to suppress lipolysis in an insulin resistant state compared to an insulin sensitive condition [40]. This is supported by the lack of changes in plasma lipids in OLETF after CR, even though we previously showed that CR improved IRI in OLETF suggesting that this improvement is largely driven by improvements in glucose metabolism independent of improved lipid metabolism. While an association between long-chain fatty acids and insulin sensitivity in obese, insulin resistant patients has been shown [41] that does not seem to apply during the early onset of insulin resistance in our model.

Proteolysis is more profound in OLETF than LETO, and is reversed after mass recovery
Normally, during negative caloric imbalance (i.e. CR) most of the amino acids generated after CR are either used by the liver for gluconeogenesis or by the muscle for energy production [42]. Consistent with these findings, we observed small but significant increases in several glucogenic amino acids, which translated into an overall increase in total amino acid concentration after CR in OLETF that returned to baseline levels after PR, but not in LETO. This increase in total amino acids suggests that lean tissue catabolism was increased and that these amino acids may be shuttled into gluconeogenesis to help support the changes in energetic demands. One advantage of this level of CR, however, is that the increased availability of cysteine may increase the production of glutathione [43], thus reducing the overall oxidative damage caused by the insulin resistant state. The potential benefit of PR during an insulin resistant condition is that carbohydrates and lipids return as the primary sources of energy, sparing lean tissue via a reduction in proteolysis. However, the potential detriment of the shift in metabolism is the partial reversal of gluconeogenesis, which could translate into exacerbation of the hyperglycemia in insulin resistant OLETF demonstrating the risk of mass recovery after an acute caloric restriction independent of recovery of visceral adiposity. Furthermore, this highlights the importance of properly managing a caloric restriction regimen following the onset of insulin resistance to ameliorate inappropriately elevated lean tissue catabolism and the potential consequences of this shift in metabolism including an increase nitrogen load on the kidneys [44] and inappropriate muscle loss. These data suggest that a reduction in adipose per se may not be sufficient in the long term to improve IRI in an insulin resistant state when gluconeogenesis is inappropriately elevated. In summary, severe acute caloric restriction was sufficient to elicit significant changes in several metabolic pathways in OLETF but not LETO, and a partial mass recovery comprised mostly by lean tissue reverted most of these changes to baseline levels with the exception of increased gluconeogenesis, and decreased lipolysis and amino acid catabolism. These shifts in substrate metabolism suggest that failure to adhere to this regimen could be detrimental in the long term due to the ability of metabolic shifts to adapt to changes in caloric intake.