Vitamin C modulates the metabolic and cytokine profiles, alleviates hepatic endoplasmic reticulum stress, and increases the life span of Gulo−/− mice

Suboptimal intake of dietary vitamin C (ascorbate) increases the risk of several chronic diseases but the exact metabolic pathways affected are still unknown. In this study, we examined the metabolic profile of mice lacking the enzyme gulonolactone oxidase (Gulo) required for the biosynthesis of ascorbate. Gulo−/− mice were supplemented with 0%, 0.01%, and 0.4% ascorbate (w/v) in drinking water and serum was collected for metabolite measurements by targeted mass spectrometry. We also quantified 42 serum cytokines and examined the levels of different stress markers in liver. The metabolic profiles of Gulo−/− mice treated with ascorbate were different from untreated Gulo−/− and normal wild type mice. The cytokine profiles of Gulo−/− mice, in return, overlapped the profile of wild type animals upon 0.01% or 0.4% vitamin C supplementation. The life span of Gulo−/− mice increased with the amount of ascorbate in drinking water. It also correlated significantly with the ratios of serum arginine/lysine, tyrosine/phenylalanine, and the ratio of specific species of saturated/unsaturated phosphatidylcholines. Finally, levels of hepatic phosphorylated endoplasmic reticulum associated stress markers IRE1α and eIF2α correlated inversely with serum ascorbate and life span suggesting that vitamin C modulates endoplasmic reticulum stress response and longevity in Gulo−/− mice.


INTRODUCTION
Maintaining adequate vitamin C (ascorbate) levels in tissues is essential for normal body function and optimal health [1]. Most mammals are capable of synthesizing their own ascorbate and are thus not prone to develop vitamin C deficiency. Humans, however, have a mutation in the gene encoding the enzyme gulonolactone oxidase (GULO) necessary for the last step of ascorbic acid synthesis [2]. Hence, humans rely entirely on dietary sources to obtain adequate amounts of

Research Paper
ascorbate. Ascorbate is a cofactor in several enzymatic reactions, including collagen synthesis that, when dysfunctional, cause scurvy [3]. Although scurvy is now considered a rare disease, epidemiological studies suggest that large subpopulations (between 5% and 30% depending on socioeconomic status, smoking status, and age) can be diagnosed with hypovitaminosis C [3][4][5][6]. Even though a hypovitaminosis C condition may not lead to scurvy, it places an individual at higher risk for metabolic abnormalities, cardiovascular diseases or cancer [3,[7][8][9].
Clinically relevant animal models of ascorbate synthesis deficiency are essential for improving our understanding of the role of vitamin C in the pathogenesis of complex diseases as well as evaluating the therapeutic potential and risks of its supplementation [10]. The progress of gene knockout mice has potentiated these areas of research. In this context, a Gulo -/mouse was created by deleting exons 3 and 4 from the Gulo gene in a C57BL/6 background [11]. This exonal knockout inactivates the enzyme and ascorbate supplementation is required to maintain viability in these mice [11]. Additional studies on this knockout model indicated elevated oxidative stress and sensorimotor deficits as well as behavioral and monoamine changes following severe ascorbate deficiency [12][13][14][15]. Furthermore, it has recently been observed that ascorbate prevents stress-induced damage on the heart through the reduction of reactive oxygen species production in Gulo -/mice depleted of ascorbate [16]. Metabolic profiling of ascorbate deficiency in Gulo -/mice using proton NMR spectroscopy revealed changes in carnitine and glutathione synthesis as well as changes in glycerophospholipid metabolism [17]. In this study, we have extended this research by measuring amino acids, biogenic acids, acylcarnitines, lysophosphatidylcholines, glycerophosphatidylcholines, sphingomyelins, and prostaglandins in the serum of Gulo -/mice treated with different concentrations of ascorbate in drinking water by targeted mass spectrometry analysis. We also examined the levels of several inflammatory cytokines, metabolic hormones, and markers of cardiovascular diseases in these animals and determined the molecules that significantly correlated with life span. We found that the median life span of Gulo -/mice increased with the amount of ascorbate in drinking water. Although the average size of mitochondria in the liver of ascorbate depleted Gulo -/mice was increased, overall mitochondrial alterations did not correlate with median life span in our mouse cohorts. In contrast, the amount of phosphorylated endoplasmic reticulum associated stress markers inositol-requiring kinase 1α (IRE1α) and the eukaryotic translation initiation factor 2α (eIF2α) in the hepatic tissues of Gulo -/mice treated with different concentrations of ascorbate significantly correlated inversely with serum ascorbate levels and median life span.

Impact of vitamin C (ascorbate) on life span and body weight of Gulo -/mice
We first investigated the effect of ascorbate supplementation on the median life span of Gulo -/mice.
Generally, four to six weeks after removal of ascorbate in drinking water, Gulo -/mice lost more than 20% of their total body weight, looked moribund, and had to be euthanized. In contrast, increasing ascorbate concentration in drinking water improved the survival of these mutant mice in a dose dependent manner ( Figure 1A). Gulo -/mice supplemented with 0.005% of ascorbate (w/v) at weaning rapidly became sick and moribund within four weeks of treatment (data not shown). Therefore this concentration of ascorbate was not used for further analysis. The median life span of Gulo -/mice treated with 0.01% ascorbate was 8 ½ months and the maximum life span was 16 months. All of these animals lost more than 20% of their total body weight, became moribund at some point during this 16 months period and had to be euthanized. Supplementation of ascorbate up to 0.4% (w/v) significantly increased the median life span to 23 months, which was close to the median life span of wild type mice (23.8 months). The maximum life span Gulo -/mice treated with 0.4% ascorbate was ~32 months compared to ~30 months for untreated wild type animals. The illnesses associated with old Gulo -/mice treated with 0.4% vitamin C included either hepatocarcinomas or myeloid leukemias. Several mice were also found dead in the cages in the morning with no sign of distress in the previous days (undetermined cause of death due to tissue autolysis). Overall, these results indicated that the life span of Gulo -/mice increased with ascorbate concentration in drinking water. The amount of ascorbate required to achieve a wild type normal life span (log rank test P > 0.05) was reached with 0.4% ascorbate in drinking water ( Figure  1A). As we have previously found that 0.4% ascorbate did not increase the life span of wild type animals [18], we did not pursue analyses with higher concentrations of ascorbate in the drinking water of Gulo -/mice.
Next we measured the amount of ascorbate in the serum of mice from each treatment group (n=5). As indicated in Figure 1B, minimal ascorbate could be detected in the serum of ascorbate-depleted Gulo -/mice (0% of ascorbate in drinking water for one month). There was a 100-fold significant increase in serum ascorbate in Gulo -/mice treated with 0.01% ascorbate compared to Gulo -/mice that had been depleted of ascorbate for one month (P < 0.01). Furthermore, there was a 5.4-fold increase when Gulo -/mice were supplemented with 0.4% ascorbate compared to 0.01% ascorbate treated mice (P < 0.01). Gulo -/mice treated with 0.4% ascorbate had similar levels of serum ascorbate compared to wild type untreated animals ( Figure 1B).
There was a significant effect of ascorbate supplementation on body weight in the Gulo -/animals.

www.impactaging.com
As indicated in Figure 1C, Gulo -/mice without ascorbate treatment were significantly leaner than Gulo -/mice treated with 0.01% or 0.4% ascorbate. The mean total body weight of Gulo -/mice treated with 0.01% or 0.4% ascorbate at four months of age were not significantly different from age-matched wild type animals. Analysis of the four-month old animals re-vealed that Gulo -/mice depleted of ascorbate for one month had significantly less visceral fat in proportion to total body weight compared to all Gulo -/mice treated with ascorbate ( Figure 1D). The weight of visceral fat in ascorbate treated Gulo -/mice (0.01% or 0.4% in drinking water) was not significantly different from untreated wild type animals ( Figure 1D).  www.impactaging.com The spleen, kidneys, heart, and liver were also weighed (Supplementary Figure S1). The spleen tended to be larger in all the Gulo -/mice compared to age-matched four-month old wild type animals. However, only the Gulo -/mice treated with 0.01% ascorbate showed a significant difference compared to wild type animals (Supplementary Figure S1A). The weight of the kidneys in Gulo -/mice depleted of ascorbate was increased (~1.2-fold) compared to age-matched wild type but was significantly decreased in Gulo -/mutant mice treated with 0.01% or 0.4% ascorbate (Supplementary Figure  S1B). The weights of the cardiac and hepatic tissues were not significantly different between cohorts (Supplementary Figure S1C and D).
Finally, Gulo -/mice depleted of ascorbate for one month ate and drank less than wild type (based on student t-test: P < 0.05; supplementary Figure S1E and F).  Effect of ascorbate on the metabolic profile of Gulo -/mice We next measured 203 metabolites employing a targeted mass spectrometry approach in the serum of Gulo -/mice treated with different amounts of ascorbate at four months of age. (Full biochemical names are provided in Supplementary Table S1). Untreated agematched wild type animals were used as the reference control. Serum metabolite concentrations in six animals of each group are shown in the supplementary Table S2.
To identify the metabolites significantly altered in either wild type, Gulo -/mice treated with 0%, 0.01%, or 0.4% ascorbate, a nonparametric Kruskal-Wallis test was applied to the data. A summary of the metabolites significantly altered (P < 0.01) in at least one of the groups is given in a form of a heatmap in Figure 2.
Changes were indicated as a Z-score value for each metabolite in the serum of each animal. The heatmap revealed 61 metabolites that significantly differed bet-ween groups ( Figure 2). These included 39 phosphatidylcholines, seven lysophosphatidylcholines, two acylcarnitines, five biogenic amines, three prostaglandins, three sphingomyelins, hexoses, and the amino acid glycine. The dendrogram on top of the heatmap indicated that the Gulo -/mice depleted of ascorbate clustered together except for one mouse (mouse identified as GNV.3). The wild type animals also formed a cluster. Gulo -/mice supplemented with either 0.01% or 0.4% ascorbate in their drinking water formed another cluster. We further explored the metabolic profile of each group by employing principal component analysis (PCA in Figure 3). The PCA approach showed a good clustering of Gulo -/mice supplemented with either 0.01% or 0.4% ascorbate in the lower right quadrant of the graph. Five Gulo -/mice depleted of ascorbate were found in the lower left quadrant (and one animal in the lower right quadrant). All wild type animals were found in the upper quadrants of the graph and did not overlap with any mouse of the Gulo -/genotype. The PCA analysis indicated that although there was a larger variability within the ascorbate depleted Gulo -/mouse group, they were localized at great distance from the wild type mice in the graph (Figure 3). The clustering of Gulo -/mice supplemented with either 0.01% or 0.4% ascorbate suggested similar metabolic profiles at four months of age. Such profiles were different from both wild type and ascorbate depleted Gulo -/mice. Finally, even though Gulo -/mice supplemented with 0.4% ascorbate had a life span and serum ascorbate levels similar to wild type animals ( Figure 1A and B), their metabolic profile did not significantly overlap with any wild type individual ( Figure 3).

Effect of ascorbate on specific metabolites in Gulo -/mice
The levels of phosphatidylcholines, lysophosphatidylcholines, sphingomyelins, and hexoses were generally lower in the ascorbate deficient Gulo -/mice compared to wild type mice. Supplementation with ascorbate increased these levels equal to or above the wild type levels ( Figure 2 and supplementary Table S2). The 15S-HETE (15(S)-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid) and 6-keto-PGF1a (6-keto prostaglandin F1α) were significantly up regulated in the serum of ascorbate deficient Gulo -/mice compared to wild type animals ( Figure 2 and supplementary Table S2). Supplementation of drinking water with 0.4% ascorbate decreased both serum 15S-HETE and 6-keto-PGF1a to wild type levels in Gulo -/mice ( Figure 2 and supplementary Table S2). LTB4 (leukotriene B4) was lower in ascorbate deficient Gulo -/mice compared to wild type animals. Supplementation of ascorbate in drinking water decreased serum LTB4 even further in Gulo -/mice (Figure 2 and supplementary Table S2). Four biogenic amines (spermine, spermidine, histamine, and alpha-aminoadipic acid) were significantly increased in ascorbate deficient Gulo -/mice compared to wild type animals ( Figure 2 and supplementary Table  S2). Histamine, spermine and its precursor spermidine were decreased to normal wild type levels when Gulo -/mice were treated with 0.4% ascorbate ( Figure 2 and supplementary Table S2). In contrast, the α-aminoadipic acid (alpha-AAA) remained elevated in Gulo -/mice compared to wild type animals in both the absence and presence of ascorbate in drinking water (Figure 2 and  supplementary Table S2). Finally, the amino acid glycine was significantly increased in ascorbate deficient Gulo -/mice compared to wild type animals ( Figure  2 and supplementary Table  S2). Supplementation of drinking water with 0.4% ascorbate in Gulo -/mice decreased serum glycine to wild type levels.

Effect of ascorbate on inflammatory cytokines and metabolic hormones in Gulo -/mice
As metabolite disturbances can lead to an inflammatory response or changes in cardiovascular risk factors, we measured the levels of 42 cytokines (including hormones) in the serum of our different mouse cohorts. Serum cytokine concentrations in eight males of each group are shown in the supplementary Table S3. To identify the cytokines significantly altered in either wild type, Gulo -/mice treated with 0%, 0.01%, or 0.4% ascorbate, the nonparametric Kruskal-Wallis test was applied to the data. A summary of the metabolites significantly altered (P < 0.01) in at least one of the groups is given in a form of a heatmap in Figure 4A. Changes were indicated as a Z-score value for each metabolite in the serum of each animal. The dendrogram on top of the heatmap indicated that the Gulo -/mice depleted of ascorbate tended to cluster together. The wild type animals also tended to form a cluster. We further explored the cytokine profile of each group by employing principal component analysis (PCA in Figure 4B). The PCA approach showed that wild type mice were mainly clustered in the upper left quadrant of the graph. Gulo -/mice supplemented with either 0.01% or 0.4% ascorbate were distributed in all four quadrants but their localization overlapped with the wild type animals. Ascorbate deficient Gulo -/mice were localized on the right side of the graph (except for mouse GNV.4). To summarize the PCA data, the serum cytokine profile of the ascorbate deficient Gulo -/mice was quite different from ascorbate treated Gulo -/and wild type mice.
The heatmap revealed seven cytokines that significantly differed between groups ( Figure 4A). These included three pro-inflammatory cytokines (interleukin-18, granulocyte-colony stimulating factor, and the macrophage inflammatory protein 1 beta), two metabolic hormones (leptin and C-peptide 2), and two cardiovascular risk factors (secreted E-selectin and the plasminogen activator inhibitor-1). The most marked findings were a general decrease in the inflammatory factors and sE-selectin in ascorbate deficient Gulo -/mice compared to wild type animals (summarized as graphs in the supplementary Figure S2). Ascorbate treatment increased the levels of these cytokines but to a level still below wild type animals. The hormones leptin and C-peptide 2 were significantly decreased in ascorbate depleted Gulo -/mice. Supplementation of ascorbate reversed the phenotype. Finally, the plasminogen activator inhibitor-1 (PAI-1) was significantly increased in ascorbate depleted Gulo -/mice compared to wild type animals (supplementary Figure S2). Supplementation of ascorbate in drinking water also reversed this phenotype in Gulo -/mice.

Correlation between cytokines and metabolites in the different Gulo -/cohorts
We determined whether there were any strong correlations (r > 0.99 and P < 0.01) between the cytokines and any of the metabolites that were altered in the Gulo -/cohorts. Only two cytokines passed such stringent criteria for these correlations. The serum levels of alpha-aminoadipic acid (alpha-AAA) showed an inverse correlation with serum sE-Selectin ( Figure 5A), while the levels of serum 6-keto-PGF1a was positively correlated with serum plasminogen activator inhibitor-1 (PAI-1) ( Figure 5B).

Correlation between ascorbate and the median life span of Gulo -/mice with metabolites or cytokines
We then determined which metabolites and cytokines were correlated with the median life span of our Gulo -/cohorts. Gulo -/mice without ascorbate in the drinking water lived approximately one month as indicated earlier. The median life span of wild type and Gulo -/mice treated with either 0.01% or 0.4% ascorbate was obtained from the graph in Figure 1A. As such, the median life span of Gulo -/mice correlated positively with serum ascorbate levels (r = 0.9918; P < 0.01 in Figure 5C). The molecule that showed strongest correlation with serum ascorbate level was serum resistin ( Figure 5D). Serum resistin levels were inversely correlated with serum ascorbate levels.
The molecules that were correlated to median life span included serum levels of octadecadienoylcarnitine (C18:2) and resistin ( Figure 5E and F). C18:2 and resistin exhibited an inverse correlation with the median life span in our mouse cohorts (with a r of at least -0.99 and a P < 0.01). Other metabolites and cytokines measured in the serum of our mouse cohorts did not correlate to ascorbate levels with a Pearson's correlation r value > 0.99. Although we obtained high Pearson's www.impactaging.com correlation r values between the median life span and the mean levels of serum resistin and C18:2, these molecules exhibited large standard deviations from the means within each group of mice such that differences between groups were not statistically significant (Kruskal-Wallis P-values > 0.01). In fact, no individual metabolite (with a Kruskal-Wallis P-values > 0.01) correlated with life span and/or ascorbate (even with a r value > 0.95 and P-value < 0.05). Next we examined specific ratios of metabolites in our cohorts of mice that could correlate with ascorbate levels or median life span with a r value > 0.95 and P-value < 0.05. Specific metabolite ratios in the serum of mice can provide insight into biological processes relevant to the activity and metabolism of ascorbate. Figure 6 provides metabolite ratios that showed significant difference between groups of mice (with ANOVA P-values < 0.05). Methionine-sulfoxide (Met-SO) can be generated via a two-electron-dependent mechanism and the ratio of Met-SO/Met in the serum can be regarded as a marker of oxidative stress [19]. As indicated in Figure 6A, Gulo -/mice treated with 0.01% ascorbate showed a significant increase in this ratio compared to wild type animals. Ascorbate is also known to affect the hydroxylation of phenylalanine into tyrosine [20]. Figure 6B indicates that the ratio of Tyr/Phe is significantly reduced in ascorbate depleted Gulo -/mice compared to wild type animals. Arginine is the only source for nitric oxide production, a molecule with important vasoprotective and anti-atherosclerotic properties [21]. Because lysine uses the same transport system as arginine for intracellular transport, intracellular arginine availability can potentially be affected by lysine [21]. As indicated in Figure 6C, the Arg/Lys ratio is significantly diminished in vitamin C www.impactaging.com depleted and in Gulo -/mice treated with 0.01% ascorbate. Hepatic arginase will also affect the levels of arginine by catabolizing it into ornithine [21]. As indicated in Figure 6D, the Arg/Orn ratio is significantly decreased in ascorbate depleted Gulo -/mice compared to wild type animals. Acylcarnitines are important molecules for mitochondrial function and lipid metabolism. Two acylcarnitine species were significantly altered in our mouse cohorts, namely C4:1 and C16-OH (based on Figure 2). Although levels of C4:1 and C16-OH did not correlate significantly with ascorbate levels or median life span, we examined the ratio of carnitine over these two acylcarnitine species. As indicated in Figure 6E and F, the C0/C4:1 and C0/C16-OH ratios were reduced in all Gulo -/mice. Finally, we examined the ratio of saturated/unsaturated phosphatidylcholine species. We pooled the serum concentrations of unsaturated phosphatidylcholines together and saturated phosphatidylcholines together to calculate the ratio. (For example, the concentrations of PC aa C32:0, PC aa C36:0, PC ae C36:0, PC ae C40:0, and PC ae C42:0 that are significantly changed in our mouse cohorts based on Figure 2 were added together to obtain the numerator of the ratio). As indicated in the Figure 6G, the ratio of saturated/unsaturated phosphatidylcholines was significantly increased in Gulo -/mice treated with 0.4% ascorbate compared to ascorbate depleted Gulo -/mice. Interestingly, among the different species of lipids altered in our mouse cohorts (Figure 2), the specific ratio PC aa C36:0/PC aa C36:2 showed a significant decrease in ascorbate depleted Gulo -/mice that was completely reversed by 0.4% ascorbate ( Figure  6H). We calculated the Pearson's correlation r values to identify the metabolite ratios that correlated significantly with serum ascorbate levels and/or the median life span of our mice. As indicated in Table 1, the ratios of Arg/Lys, Tyr/Phe, saturated/unsaturated lipids, and PC aa C36:0/PC aa C36:2 correlated significantly with median life span. Serum ascorbate levels correlated significantly with the Arg/Lys, Tyr/Phe, and PC aa C36:0/PC aa C36:2 ratios.

No correlation between the median life span of Gulo -/mice and body weight
Since body mass and visceral fat may have an effect on life span, we determined whether they were significantly correlated in our mouse cohorts. As expected, visceral fat wet weight correlated significantly with total body weight (r = 0.9972; P < 0.01) in these mice. The median life span of Gulo -/mice, however, did not correlate with the mean total body weight of mice or the visceral wet weight/body weight ratio ( Table 2). The median life span did not correlate with the weight of the other organs. Similarly, serum ascorbate levels did not significantly correlate with total body weight, visceral fat weight or the weight of any organ analyzed ( Table 2).
Visceral fat wet weight (and thus total body weight) significantly correlated with the levels of several serum phosphatidylcholine lipid species and serum hexoses (supplementary Table S4). Such phosphatidylcholine lipid species and hexoses, however, did not correlate significantly with the median life span of our mice (Pvalues > 0.05). www.impactaging.com The liver plays a pivotal role in nutrient, hormone, lipid, and metabolic waste product processing, thereby maintaining body homeostasis [22]. It is also the normal site of ascorbate synthesis in mice [2]. Since ascorbate treated Gulo -/mice exhibited alterations in several lipid species, we next measured several cellular morphological parameters in the hepatic tissue. We first examined sinusoidal endothelial fenestration in liver of mice. Liver endothelial defenestration is a recognized age-related change [23]. Scanning electron microscopy revealed no significant difference in the number of sinusoidal endothelial fenestration between our cohorts of mice (supplementary Figure S3A). In addition, fenestration diameter was not significantly different between cohorts (supplementary Figure S3B and C).
Since the mitochondrion is important in metabolism, we next examined the morphology of mitochondria in the liver of our mouse cohorts. Enlarged mitochondria in liver tissues have been reported in a mouse model of premature aging [24] and in human patients with liver disease [25]. Transmission electron microscopy revealed a significant increase in the dimension of hepatic mitochondria in ascorbate depleted Gulo -/mice compared to wild type animals ( Figure 7A). Supplementation of ascorbate in drinking water (0.01% or 0.4%) significantly decreased the size of the mitochondria to wild type dimensions. In contrast, the surface density of the mitochondrial envelop was significantly lower in all Gulo -/mice (with 0%, 0.01% or 0.4% ascorbate) compared to wild type animals ( Figure 7B; Tukey post-ANOVA test; P < 0.01). Finally, the number of mitochondria/hepatic surface area was not significantly different between cohorts of mice (supplementary Figure S4A and B; Tukey post-ANOVA test; P > 0.05). The levels of reactive oxygen species (ROS) were measured in the liver of mice. As indicated in Figure 7C, ROS level was increased in ascorbate depleted and 0.01% ascorbate treated Gulo -/mice compared to wild type animals (Tukey post- ANOVA test; P < 0.05). ROS level was decreased in Gulo -/mice treated with 0.4% ascorbate to wild type levels. Liver ROS levels significantly correlated inversely with serum ascorbate levels ( Figure 7D).
We determined whether there were significant correlations between cytokines or metabolites and mitochondrial morphology in our mouse cohorts. Figure  7 (panels E-H) gives the molecules with the highest correlations (r > 0.99 and P < 0.01). The prostaglandin derivative 15S-HETE and the secreted protease PAI-1 were positively correlated with hepatic mitochondrial size ( Figure 7E and F). Serum sE-selectin levels in return, correlated positively with the surface density mitochondrial envelop area ( Figure 7G). In contrast, serum alpha-AAA levels correlated negatively with mitochondrial surface density envelop area ( Figure 7H). Finally, hepatic mitochondrial size also significantly (r > 0.95; P < 0.05) correlated positively with histamine, spermidine, spermine, and 6-keto-PGF1a, and inversely with body weight, wet visceral fat weight, PC aa C42:4, and PC aa C42:6 ( Table 3). The surface density mitochondrial envelop area significantly correlated with the ratio of C0/C4:1 and inversely with serum C16-OH levels and the Met-SO/Met ratio ( Table 3).

Effect of ascorbate on stress markers in the liver of Gulo -/mice
Since we could detect changes in serum inflammatory cytokines, lipids, and hepatic ROS in our different cohorts of mice, we analyzed the levels of several stress markers in the liver of these mice. Phosphorylation of IRE1α, PERK, and eIF2α were first analyzed by western blotting (Figure 8A). IRE1α acts as an endoplasmic reticulum (ER) stress sensor and is part of the unfolded protein stress response [26]. As indicated in Figure 8A and B, the levels of phosphorylated IRE1α protein were increased in the liver of Gulo -/mice treated with 0.01% or depleted of ascorbate compared to wild type animals. Supplementation of 0.4% ascorbate in drinking water significantly reduced the levels of phosphorylated IRE1α proteins compared to ascorbate depleted Gulo -/mice ( Figure 8B; shown only by student t-test: P < 0.05). Total IRE1α, in return, was significantly increased in 0.4% ascorbate treated Gulo -/mice compared to wild type, ascorbate depleted and 0.01% ascorbate treated Gulo -/mice ( Figure 8C; Tukey post-ANOVA test: P > 0.01). PERK is a transmembrane protein located in the ER and contains a kinase domain activated upon ER stress that phosphorylates eIF2α,  www.impactaging.com thus inhibiting the translation of mRNAs [27]. ER stress also leads to PERK autophosphorylation. As indicated in Figure 8A and D, the levels of phosphorylated PERK was significantly increased in ascorbate depleted Gulo -/mice compared to wild type animals (Tukey post-ANOVA or student t-test: P < 0.05). Total PERK levels (D) Ratio of phosphorylated PERK signal over β-actin signal from the western blots. (Tukey post ANOVA test: *P < 0.05 compared to ascorbate depleted Gulo -/mice). (E) Ratio of total PERK signal over β-actin signal from the western blots. (Tukey post ANOVA test: *P < 0.05 and **P < 0.01 compared to type mice). (F) Ratio of phosphorylated eIF2α signal over β-actin signal from the western blots.
www.impactaging.com were significantly increased in all Gulo -/mice compared to wild type animals ( Figure 8E; Tukey post-ANOVA or student t-test: P < 0.05). Phosphorylated eIF2α was significantly decreased in 0.4% ascorbate treated Gulo -/mice compared to ascorbate depleted Gulo -/mice ( Figure  8A and F). Total eIF2α level was significantly decreased in 0.01% ascorbate treated Gulo -/mice compared to wild type mice (based on student t-test: P < 0.05; Figure 8G).
GRP78, also referred as the immunoglobulin heavy chain-binding protein (BiP), is a member of the heatshock protein-70 family and is involved in the folding and assembly of proteins in the ER [26]. There was a decrease in GRP78 in Gulo -/mice treated with 0.4% ascorbate ( Figure 8A), but overall there was no significant difference in the levels of this protein between groups of mice ( Figure 8H; Tukey post-ANOVA or student t-test: P > 0.05).
The NFκB transcription factor determines cell response to a wide variety of stresses, including inflammation, and was recently shown to be one of the most strongly age-associated markers [28]. As indicated in Figure 9A and B, phosphorylated NFκB levels were significantly different between groups of mice (Tukey post-ANOVA test: P > 0.01). Phosphorylation (and thus activation of NFκB) was increased in Gulo -/mice treated with 0.4% ascorbate compared to wild type and Gulo -/mice treated with 0.01% ascorbate. Total NFκB levels were similar in all groups of mice ( Figure 9A and C).
We also examined the levels of total and phosphorylated p38 MAP kinase in the hepatic tissues of our mouse cohorts. As indicated in Figure 9D and E, the addition of ascorbate in drinking water decreased total and phosphorylated levels of p38 in the liver of Gulo -/mice compared to wild type animals (Tukey post- post-ANOVA; P < 0.05). However, total and phosphorylated levels of p38 were not higher in the liver of ascorbate depleted Gulo -/mice compared to wild type animals.

Levels of phosphorylated IRE1α and eIF2α correlated with levels of ascorbate and median life span
We determined whether there were significant correlations between the stress markers analyzed in our mouse cohorts and the median life span or the serum ascorbate levels. Table 4 provides a summary of the correlations. Phosphorylated IRE1α and phosphorylated eIF2α were the only markers that correlated significantly (and inversely) with the median life span of mice in our different cohorts with a P-value of at least < 0.05. They also significantly correlated inversely with serum ascorbate levels ( Table 4). Phosphorylated or total PERK, GRP78, NFκB, and p38 MAP kinase did not correlate significantly with median life span or ascorbate levels (Table 4). Thus, different phosphorylated markers of ER stress response correlated significantly with the median life span of Gulo -/mice treated with different concentrations of ascorbate in drinking water.

Correlation between stress markers and metabolites or cytokines
We next looked at the significantly altered metabolites or cytokines (Figures 2 and 4A) that correlated (with a r > 0.95 and a P < 0.05) with the different stress markers analyzed in the liver of our mouse cohorts. Significant correlations with the ER stress markers are shown in Table 5. Importantly, phosphorylated IRE1α and eIF2α correlated positively with the ratio of saturated to unsaturated phosphatidylcholines. Although the phosphorylation of PERK did not correlate significantly with median life span (only a tendency with a r = -0.9196), it positively correlated with the phosphorylation of eIF2α proteins. Total PERK inversely correlated with the surface density of mitochondrial envelop area (Table 5) and with several of the molecules that correlated with this mitochondrial parameter (like sE-selectin, C16-OH, alpha-AAA, and the Met-SO/Met ratio; compare Tables 3 and 5). The amount of the ER stress marker GRP78 correlated inversely with total IRE1α levels in the liver of our mouse cohorts. Finally, the levels of phosphorylated and total eIF2α correlated inversely with several lysophosphatidylcholine and phosphatidylcholine species in the serum of our mice (Table 5).
Total and phosphorylated p38 correlated inversely with several phosphatidylcholine species including lipids with very long carbon chains in our different cohorts of mice (Table 6). Phosphorylated p38 also significantly correlated positively with total p38 in the liver of mice. Finally, phosphorylated NFκB did not significantly correlate with metabolites or cytokines.

DISCUSSION
The anti-oxidant and anti-inflammatory properties of ascorbate have been inversely correlated with several chronic diseases [29]. The impact of ascorbate on overall life span, however, has been inconsistent from one study to another [30][31][32]. Various confounding factors have been suggested to explain these discrepancies including sample size, dose duration, genetic variation, or disease status [33]. Despite the controversies surrounding the therapeutic potential of ascorbate in various chronic diseases, it is becoming clear that insufficient dietary vitamin C places individuals at greater risk for heart disease, cancer, or degenerative conditions [5,8,34,35]. It has been reported that in the United States smokers and lowincome persons suffer from varying degrees of ascorbate deficiency [5,7]. Low dietary vitamin C may not necessarily lead to scurvy but suboptimal levels will impair the normal function of enzymes in several subcellular compartments like the nucleus, the endoplasmic reticulum, or the mitochondrion [36,37]. Studies of the pathogenesis of human diseases related to vitamin C deficiency has benefited from the availability of mutant mice lacking the enzyme responsible for the last step of ascorbate synthesis [11]. In this study, we analyzed the impact of different concentrations of ascorbate on the serum levels of 203 metabolites by mass spectrometry and the serum concentrations of 42 cytokines involved in inflammation, cardiovascular diseases, or lipid metabolism in such mutant mice.
The median life span of Gulo -/mice increased with the amount of ascorbate supplemented in the drinking water. Concomitantly, serum ascorbate levels in Gulo -/mice increased with the amount of ascorbate in the diet. In fact, the median life span of the different cohorts of mice (including untreated wild type mice) correlated positively with the mean levels of ascorbate measured in the serum of each group of mice. Although we studied small cohorts of mice, it was clear that the maximum life span of Gulo -/mice treated with 0.4% ascorbate was similar to untreated wild type animals. Higher concentrations of ascorbate were not tested in this study as we previously found that wild type mice treated with 0.4% of ascorbate did not exhibit an increase in their longevity compared to untreated wild type animals [18]. The absence of ascorbate in drinking water lead to a severe body weight loss in the Gulo -/mouse as described previously [11,17] with a concomitant decrease in visceral fat weight. This decrease in body mass is likely due to the observed decrease in food intake [15,38]. Importantly, the decrease levels of hexoses and of most lysophosphatidylcholine and phosphatidylcholine molecules in the serum of ascorbate deficient Gulo -/mice compared to age matched four-month old wild type animals reflected this phenotype. Supplementation of 0.01% or 0.4% ascorbate in drinking water increased the levels of these molecules to wild type levels or even above wild type concentrations, although ascorbate treated Gulo -/mice were not significantly overweight compared to wild type animals ( Figure 1C). The metabolic hormones Cpeptide 2 and leptin were also significantly decreased in ascorbate deficient Gulo -/mice (supplementary Figure  S2). Supplementation of 0.01% or 0.4% ascorbate reversed such phenotype in Gulo -/mice. However, the median life span and ascorbate levels did not significantly correlate with body weight, visceral fat weight, or with these two metabolic hormones. Apart from ascorbate, the only metabolite that highly correlated (P < 0.01) with the median life span of our cohorts under study was octadecadienoylcarnitine C18:2 (r = -0.9969). Octadecadienoylcarnitine C18:2 (or the linoleyl carnitine) is a long-chain acyl fatty acid derivative ester of carnitine. Long-chain acyl fatty acid derivatives are known to accumulate in the cytosol and serum of patients suffering from mitochondrial carnitine palmitoyltransferase II deficiency, the most common inherited disorder of lipid metabolism in adults [39]. Octadecadienoylcarnitine is also known to inhibit the mitochondrial complex IV resulting in an increase in reactive oxygen species [40]. Accordingly, Gulo -/mice with low levels of serum ascorbate exhibited increased ROS in their liver compared to age-matched wild type animals.
Resistin was the only cytokine significantly correlating (P < 0.01) with the median life span of mice in our cohorts (r = -0.9912). Interestingly, resistin also correlated with serum ascorbate levels in our different cohorts of mice ( Figure 5D). Resistin is a cytokine produced by fat and immune cells that can modulate the obesity pro-inflammatory environment [41]. Resistin has been described as a cytokine that links obesity to diabetes in mice [42]. Importantly, it has been reported that ascorbate supplementation in humans significantly reduces resistin levels, presumably through its antioxidant activity, and this is independent of changes in inflammatory or other metabolic variables [43]. The results with our Gulo -/cohorts recapitulate the data from this report. Finally, low resistin levels have been associated with low prevalence of metabolic syndrome and healthier centenarians [44]. Thus, a low serum resistin level in ascorbate treated Gulo -/mice would be in agreement with a longer life span. However, despite these interesting observations in our mouse cohorts, the difference in serum C18:2 and resistin levels between groups of mice was not statistically significant based on Kruskal-Wallis analyses.
We also examined the ratios of different metabolites and these showed significant differences between our groups of mice and identified potential metabolic pathways significantly correlated with median life span. Eight specific ratios were significantly different between our groups of mice (based on ANOVA analyses, Figure 6). Four metabolite ratios correlated significantly with the median life span ( Table 1). The Tyr/Phe ratio correlated positively with median life span. Hydroxylation of phenylalanine to tyrosine is a pathway affected by ascorbate concentrations in vivo [20]. Accordingly, we found a significant correlation between Tyr/Phe ratio and serum ascorbate levels. The Arg/Lys ratio also correlated positively with median life span. Arginine is required for nitric oxide synthesis, a molecule with cardiovascular protective properties. Since lysine competes with arginine for the same intracellular transporter [21], this ratio provides an indirect assessment of arginine bioavailability for nitric oxide synthesis and cardioprotection. In this context, the low Arg/Lys ratio in ascorbate depleted Gulo -/mice is in agreement with the aortic wall damage observed in these mice when treated with low levels of ascorbate [11].
Interestingly, the ratio of saturated/unsaturated phosphatidylcholines and more specifically the ratio of PC aa C36:0/PC aa C36:2 phosphatidylcholines correlated significantly with median life span. The life span of Gulo -/mice treated with different concentrations of ascorbate did not correlate with body weight or visceral wet weight. This indicates that in Gulo -/mice the ratio of saturated/unsaturated lipids is a more relevant determinant of life span than total serum lipids.
We observed that the dimensions of hepatic mitochondria increased in ascorbate depleted Gulo -/mice. Supplementation of 0.01% ascorbate was enough to reverse mitochondrial dimensions to wild type size ( Figure 7). Hepatic mitochondrial size correlated significantly with serum levels of 15S-HETE and PAI-1. Interestingly, it has been reported that PAI-1 levels can regulate mitochondrial mass in cancer cells [45]. The positive correlation between PAI-1 and mitochondrial dimension in liver of ascorbate depleted www.impactaging.com Gulo -/mice is consistent with such findings. The impact of 15S-HETE on mitochondrial size is unknown. However, 15S-HETE possesses anti-proliferative properties. Increased serum levels of 15S-HETE may reflect a response to the abnormal mitochondrial morphological dimension and increased ROS observed in ascorbate depleted Gulo -/mice. Mitochondrial dimension also correlated positively with histamine, spermine, and spermidine. In contrast, hepatic mitochondrial size correlated inversely with specific long chain phosphatidylcholines (PC aa C42:4 and PC aa C42:6), body weight and visceral fat weight. Interestingly, it has recently been reported that excess visceral adiposity is significantly associated with mitochondrial size and dysfunction [46]. Spermine and spermidine are known to modulate hepatic mitochondrial metabolism [47]. Histamine, in return, has been shown to affect the size of mitochondria of gastric cells in starved guinea pigs [48]. Thus, an increase of serum alpha-AAA may represent an impaired turnover of decarboxylation 2-oxoadipate to glutaryl-CoA, which is the last step in the lysine degradation pathway [49]. Aging, diabetes, sepsis, and renal failure are known to catalyze the oxidation of lysyl residues to alpha-AAA in human skin collagen and potentially other tissues. Proteolytic breakdown of these tissues can lead to the release of free alpha-AAA [50]. Dysfunctional endothelial cells leads to elevated levels of alpha-AAA, which is also thought to be a sign of lysyl residues breakdown in proteins through oxidative stress and reactive oxygen species [51]. We also observed a significant inverse correlation between serum alpha-AAA and serum sE-selectin ( Figure 5). Eselectin is expressed on cytokine-activated endothelial cells and contributes to the adhesion of leukocytes to the endothelium and activation of the immune cells.
After cell activation, E-selectin is eliminated from the cytoplasmic membrane by shedding into the circulation as secreted E-selectin (or sE-selectin). It is possible that in ascorbate depleted Gulo -/mice, cell damage reflected by the increased serum alpha-AAA and abnormal surface density of mitochondrial envelop area leads to sE-selectin response required for the elimination of irreversibly damaged cells. Injection of these molecules in Gulo -/mice will be required to determine the cause from effect.
Addition of ascorbate in the diet reversed the decreased life span of Gulo -/mice. However, even though Gulo -/mice treated with 0.4% ascorbate had a life span similar to wild type animals, their overall metabolic profile was still different from the wild type profile. The PCA analysis clearly indicated that Gulo -/mice treated with either 0.01% or 0.4% ascorbate clustered together and did not overlap with the wild type animals ( Figure 3). Thus, ascorbate did not re-establish a complete wild type metabolic profile. In contrast, the cytokine profile of Gulo -/mice treated with 0.4% ascorbate overlapped with that of the wild type animals ( Figure 4B).
Although we could detect changes in the levels or the activation of different stress markers in the liver of ascorbate depleted Gulo -/mice, the phosphorylated ER stress associated markers IRE1α and eIF2α were the only proteins that highly correlated (and inversely) to the mean life span of Gulo -/mice treated with different concentrations of ascorbate. Phosphorylated IRE1α and eIF2α proteins also inversely correlated significantly with serum ascorbate levels in our different mouse cohorts, suggesting that the ER stress can be alleviated by ascorbate. Accordingly, it has been reported that ascorbate protects against metal-induced ER stress in male gonads of mice [52]. Phosphorylated IRE1α also significantly correlated inversely with the Tyr/Phe and saturated/unsaturated phosphatidylcholine ratios in our mouse cohorts. Unbalanced fatty acid ratio is known to affect membrane lipid composition [53] and may thus affect the ER membrane, trans-membrane transporters, or membrane associated enzymatic activities in Gulo -/mice. Phosphorylation and thus activation of PERK during ER stress leads to eIF2α phosphorylation and inhibition of mRNA transcription. As expected, we found a significant positive correlation between activation of PERK and phosphorylation of eIF2α. However, unlike the phosphorylation of eIF2α, we did not detect a significant correlation (only a tendency) between phospho-PERK levels and the median life span of our mouse cohorts. Importantly, phosphorylation of eIF2α can occur independently of PERK activity in cells [54] or tissues upon inflammation or oxidative stress [55], a possibility that will need further investigation in our aging mouse cohorts. Interestingly, total PERK levels correlated inversely with the surface density of mitochondrial envelop area in our mouse cohort. It has been reported that PERK is an essential component of the mitochondria-associated ER www.impactaging.com membranes that establish a physical and functional connection between the ER and the mitochondria. PERK serves also as a structural tether at the ERmitochondria interface regulating inter-organellar crosstalk in ROS-induced cellular response [56]. It is thus possible that Gulo -/mice increased total PERK protein levels in response to the alterations observed in the surface density of mitochondrial envelop area in our mouse cohorts. Finally, we found that other stress markers like the p38 MAP kinase and the transcription factor NFκB did not correlate with the life span of ascorbate treated Gulo -/mice.
To conclude, our study demonstrates the impact of suboptimal levels of ascorbate on longevity and the metabolic profile of a mouse model, that similar to humans, lacks the enzyme required for the synthesis of ascorbate. Vitamin C supplementation alleviated the abnormal levels of several lipid species ( Figure 2) and cardiovascular risk factors ( Figure 4 and Table 1). Finally, serum ascorbate levels and longevity correlated significantly with the ER stress response as seen by the activation of IRE1α and inhibition of eIF2α through phosphorylation reactions. As such, this study is the first work reporting a significant correlation between ER stress response and longevity in Gulo -/mice treated with different concentrations of ascorbate. Mice that lost 20% of total body weight, became immobile, or moribund were sacrificed for histological examination of their organs as described previously [57]. One cohort of Gulo -/mice was maintained on standard diet and supplemented with 0.4% of L-ascorbate (Sigma-Aldrich, Oakville, ON) (w/v) in drinking water from weaning until the age of four months. A second cohort of Gulo -/mice was maintained on standard diet and supplemented with 0.01% of L-ascorbate (w/v) in drinking water from weaning until the age of four months. A third cohort of Gulo -/mice was maintained on standard diet and supplemented with 0.01% of L-ascorbate (w/v) in drinking water from weaning until the age of three months. Ascorbate was then removed from drinking water for four weeks. Wild type control C57BL/6 mice were maintained in the same room with no ascorbate supplementation in drinking water and were used as our normal reference. The analyses were performed on four month-old animals as more than 85% of Gulo -/mice treated with 0.01% ascorbate were healthy at that age. with electrospray ionization. The experimental metabolomics measurement technique has been previously described [53]. Eicosanoids and other oxidized polyunsaturated fatty acids were extracted from samples with aqueous acetonitrile that contained deuterated internal standards. The metabolites were determined by HPLC-tandem mass spectrometry (LC-MS/MS) with Multiple Reaction Monitoring (MRM) in negative mode using a SCIEX API 4000 QTrap mass spectrometer with electrospray ionization. The LC-MS/MS method used for the analytical determination of eicosanoids has been published [58]. Accuracy of the measurements (determined with the accuracy of the calibrators) was in the normal range of the method www.impactaging.com (deviations from target ≤ 20 %) for all analytes. In total, 203 different metabolites were measured. The metabolomics data set contains 21 amino acids, 19 biogenic amines, one hexose (H1), free carnitine (C0), 40 acylcarnitines (Cx:y), hydroxylacylcarnitines (C(OH)x:y), and dicarboxylacylcarnitines (Cx:y-DC), 15 sphingomyelins (SMx:y) and N-hydroxylacyloylsphingosylphosphocholine (SM (OH)x:y), 77 phosphatidylcholines (PC, aa = diacyl, ae = acyl-alkyl), 14 lyso-phosphatidylcholines, and 17 eicosanoid acids and prostaglandins. Lipid side chain composition is abbreviated as Cx:y, where x denotes the number of carbons in the side chain and y the number of double bonds. For example, "PC ae C30:1" denotes an acylalkyl phosphatidylcholine with 30 carbons in the two fatty acid side chains and a single double bond in one of them [53]. Full biochemical names are provided in Supplementary Table S1  Morphological and ultrastructural studies of liver tissues. For liver morphology studies, liver perfusion was performed as described previously [59]. Following fixation, liver samples for transmission electron microscopy were embedded in Spurrs resin, sectioned, and examined using a Philips CM10 transmission electron microscope. Liver morphology was assessed for mitochondrial density, mitochondrial size, surface density of mitochondrial envelop using ImageJ software. Three to five mice per genotype were analyzed by electron microscopy techniques. Fifteen micrographs per animal of hepatocytes cytoplasm were taken at 9,900 × magnification. Mitochondria were manually counted and the number of mitochondria per cytoplasmic unit area was calculated. Mitochondria area was measured by tracing around five random mitochondria in each micrograph. The surface density of mitochondrial envelope (Sv) was estimated using the formula Sv=2*I/Lt where I represents the number of intersections of mitochondria envelops with parallel lines (1µm apart) and Lt the total length of lines as described [60]. Scanning electron microscopy was performed as previously described [61]. Ten images (magnification x 20,000) were taken for each animal for analysis of fenestration diameter and endothelial porosity.

MATERIALS AND METHODS
Reactive oxygen species (ROS) measurements in liver tissue. Liver lysates in RIPA buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, 30 mM NaF, 60 mM glycerophosphate, 20 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 1mM phenylmethylsulfonylfluoride and complete protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN)) were incubated with 10 µg/mL of the dye 2'-7' dichlorofluorescein diacetate (Sigma-Aldrich Canada Ltd, Oakville, ON) for 1h at 37°C. This dye is highly fluorescent upon oxidation. As control, RIPA buffer was also incubated with dichlorofluorescein diacetate and 100 µL (500 µg of liver proteins) of the samples were put into 96-well plates. Fluorescence was measured with a Fluoroskan Ascent fluorescence spectrophotometer (Thermo Electron Inc., Milford, MA). The excitation and emission wavelengths used were 485 and 527 nm, respectively. Background fluorescence was extracted from the dichlorofluorescein value for each sample and the final result was expressed as units of fluorescence per gram of proteins.