The Metabolic Origins of Mannose in Glycoproteins*

Background: Metabolic origins of mannose in glycoproteins have not been studied. Results: Glucose is the major source of N -glycan mannose, but exogenous mannose provides up to 50%. Man salvaged from glycoprotein turnover, glycogen and gluconeogenesis is insignificant. Conclusion: Exogenous mannose is incorporated into N -glycans much more efficiently than that derived from glucose. Significance: This study provides quantitation of metabolic origins of Man in N -glycans. ABSTRACT Mannose in N -glycans is derived from glucose through phosphomannose isomerase (MPI, Fru-6-P  Man-6-P) whose deficiency causes a congenital disorder of glycosylation (CDG)-Ib (MPI-CDG). Mannose supplements improve patients’ symptoms because exogenous mannose can also directly contribute to N -glycan synthesis through Man-6-P. However, the quantitative contributions of these and other potential pathways to glycosylation are still unknown. We developed a sensitive GC-MS-based method using [1,2- 13 C]glucose and [4- 13 C]mannose to measure their contribution to N -glycans synthesized under physiological conditions (5 mM glucose and 50 µ M mannose). Mannose directly provides ~10  45% of the


INTRODUCTION
Both glucose and mannose can contribute mannose to N-glycan synthesis through a common precursor, Man-6-P, which is converted to Man-1-P via phosphomannomutase (PMM2) and on to GDP-mannose. For over 35 years, biosynthetic studies of glycoproteins have relied on the labeling specificity of [2-3 H]mannose. Intracellular conversion to [2][3] H]Man-6-P then leads to two mutually exclusive fates: 1) glycan synthesis or 2) MPI-mediated catabolism to Fru-6-P with release of 3  2 Other studies in cells (2), mice (3), and humans (4,5) all show that surprisingly small increases in exogenous mannose dramatically improves deficient glycosylation (2)(3)(4)(5). Patients carrying hypomorphic, life-threatening mutations in MPI suffer from coagulopathy, protein-losing enteropathy, hypoglycemia, and liver dysfunction, but most of these abnormalities reverse when patients are given ~0.6 g mannose/kg/day as dietary supplements raising their plasma mannose concentration by 34 fold (4). Mannose also reverses insufficient glycosylation in patientderived cells (2). Mannose rescues embryonic lethality of mice carrying mutations in Pmm2 that die by mid-gestation, but when dams are given 0.9% mannose in their drinking water, pups are born and thrive (3). However patients with PMM2 mutations do not seem to respond to mannose therapy (6,7). These observations prompted us to quantitatively assess the contributions of mannose to mannose (Man M ) and glucose to mannose (Man G ) in glycosylation as well as their catabolic fate. In this report, we describe and apply new methods to metabolically label cells with stable isotopes (8) and determine the relative and quantitative contributions of mannose (Man M ) to N-glycans. We found that the contribution of exogenous mannose was previously under estimated.
We also combined stable monosaccharide isotope labeling with deuterium oxiside (D 2 O) and found that other potential sources of mannose including mannose salvaged from degraded glycoproteins (Man S ), glycogen (Man GL ) and gluconeogenesis do not make significant contributes to N-glycosylation.
Cell Lines-Normal fibroblasts were obtained from American Type Culture Collection and Coriell Cell Repositories. All the cell lines were grown in DMEM with 1 g/l glucose containing 2 mM glutamine, 10% Fetal Bovine Serum (FBS) and penicillin-streptomycin. DMEM and FBS were obtained from Mediatech and Thermo Scientific Hyclone, respectively. Dialyzed FBS was obtained from Invitrogen. Glutamine and antibiotics were purchased from Omega Scientific Inc. ReagentsBSTFA (N,O-bis[trimethylsilyl] trifluoroacetamide) was purchased from Thermo Scientific. All the other reagents and enzymes were purchased from Sigma-Aldrich.
General procedure for Stable isotope labeling and sample preparation-Cells were grown in 60 mm dish to 75% confluency, labeled in DMEM glucose free medium reconstituted with stable isotopes of glucose and mannose (5 mM and 50 µM, respectively) plus 10% dialyzed FBS for 24 h unless otherwise noted. For glycan analysis, cells were washed with PBS, extracted with chloroform methanol (2:1), deionized water and chloroform-methanol-water (10:10:3) (9). The cell pellet was dried, resuspended in PBS/1% SDS and N-glycans were released by PNGase F digestion. Purified glycans were hydrolyzed by 2 M trifluoroacetic acid at 100 °C, and dried. For glycogen analysis, cells were extracted with 85% EtOH, and the resulting pellets were dissolved in 0.1 M NaOH/0.1% sodium lauryl sulfate and digested with amyloglucosidase (10) Hydrolysates from N-glycan and digests from glycogen were derivatized as described in (11), dried, resuspended in chloroform and analyzed by GC-MS. For metabolite analysis from glycolysis, aliquot of 50 µl medium was extracted with 250 µl 50% methanol and 150 µl chloroform. The aqueous layer was separated and dried. The resulting solid was derivatized with 50 µl ethoxylamine hydrochlorides in pyridine (20 mg/ml) at 80 °C for 20 min, followed by addition of 50 µl TBDMS-TFA (N-tert-butyl-dimethylsilyl-N-methyltrifluoroacetamide) and heated at 80 °C for another 1h. Derivatized sample were centrifuged and supernatant was directly injected for GC/MS analysis (12).
Gas Chromatography/Mass SpectrometrySugar analysis was done using a 15 m  0.25 mm  0.25 μm SHRXI-5ms column (Shimadzu, Kyoto, Japan) on QP2010 Plus Gas Chromatography 3 Mass Spectrometry (GC-MS) (Shimadzu, Kyoto, Japan) with m/z range 140500 and ~1.5 kV detector sensitivity. GC-MS ion fragment intensities obtained using GC/MS Solution version 2.50 SU3 from Shimadzu Corporation. For each fragment, the intensities were corrected for the natural abundance of each element using matrixbased probabilistic methods as described (1315). The 13 C/( 12 C + 13 C) or 2 H/( 1 H + 2 H) ratios were used to calculate isotopic labeling proportion.
Choice of isotopes to determine the origin of Man in N-glycansThe following fragmentation series were used to identify the origin of mannose in Nglycan: m/z 145 and 187 for [6,6- Total N-glycan and glycogenThe amount of mannose derived from N-glycans was analyzed by GC-MS and calculated as mannosyl units/mg protein based on the mannose standard curve. Glucose derived from glycogen was also analyzed by GC-MS and calculated as difference from aminoglycosidase (AG) untreated samples. A glucose standard curve was used to calculate glycogen and its total amount was expressed as glycosyl unit per mg protein (16). Preparation for stable isotopes with D 2 O labeling or chase with D 2 O-DMEM glucose free medium with dialyzed serum was lyophilized and reconstituted in deuterium water prior to labeling. Stable isotopes and other additives were also dissolved in deuterium water prior to use. The contribution of [UL-13 C]Man/Glc to Man-6-P and Glc-6-P was calculated with the m475/(m471 + m475) ratio. The contribution of [UL-13 C]Man/Glc to Fru-6-P 4 was calculated with the m462/(m459 + m462) ratio (Fig. 6B).
Rate of sugar uptake from the mediumRates of sugar uptake were calculated based on the remaining concentration of each sugar in the cultured medium at time 024 h. Cells were grown in 5 mM glucose and 50 µM mannose, and an aliquot of 50 µl medium was extracted with 250 µl 50% methanol and 150 µl chloroform. The aqueous layer was separated and dried. The resulting solid was derivatized in the same manner as that described under sugar phosphate analysis (17). For quantitation of glucose and mannose, the total ion count peak area of the respective Z-and E-forms of the sugars, or specific mass ions, were used for calculation based on standard curves.

Choice
of Mannose and Glucose IsotopesExogenous mannose and glucose can both contribute to mannose in N-glycans, as shown in Scheme 1. We supplied cells with mannose and glucose containing stable isotopes ( 2 H or 13 C) at different positions in the sugar backbone. Monosaccharides incorporated into cellular Nglycans were released by PNGase F digestion and hydrolyzed. The released monosaccharides were converted to aldonitrile acetate derivatives for analysis by GC-MS. Mannose and glucose were baseline-separable by GC ( Fig. 2A), and each produced similar mass fragments. By varying the locations of the heavy isotope on each precursor sugar (Fig. 2B), we could distinguish their origin in N-glycans as being from Glc or Man. (Fig. 2C) (11). We tested combinations of [1,[2][3][4][5][6][7][8][9][10][11][12][13]  C]mannose as precursors. Most of the variously labeled glucose or mannose species (Man G or Man M , respectively) were incorporated into mannose in N-glycans at similar rates, suggesting little isotope effects. The single exception was [6,6-2 H]glucose, which showed a 26% reduction in incorporation into N-glycans, presumably due to the presence of two bulky deuterium labels at the 6-position which reduced either the rate of glucose transport or subsequent metabolism (data not shown). We chose [1,2-13 C]glucose, and [4-13 C]mannose for further studies since they gave the clearest signals using the mass isotopomer patterns of m/z 314 and m/z 242 aldonitrile-sugar fragments (+2 or +1 mass unit labeling, respectively). Exchanging the location of the labels between the two precursors had no effect on their relative proportions in N-glycans.
Samples containing as little as 5% incorporated heavy isotope gave reliable values above natural abundance background for each isotope after 2 h of labeling (Fig. 3A). Fibroblasts were routinely labeled for 1624 h without deleterious effects to the cells. Chasing heavy isotope-labeled cells with non-labeled medium showed that N-glycan mannose (content = 24 nmole/mg protein) has a t 1/2 of 24 h (Fig. 3B).
Fate and Distribution of MannoseUnder physiological concentrations (5 mM glucose, 50 µM mannose) several studies have shown that only a small amount of exogenous [2-3 H]mannose is incorporated into glycans (~2%) while most is released as 3 HOH (98%) (1). Using stable isotopes and non-labeled sugars, we found that fibroblasts take up exogenous mannose at a rate of 9.422 nmole/mg/h protein and glucose at 15002200 nmole/mg/h protein (Table 1). Among five different cell lines tested, glucose and mannose contribution to N-glycans ranged from 0.10.4 nmole/mg/h and 0.10.2 nmole/mg/h, respectively, showing that, relative to their uptake rates, mannose is incorporated into N-glycans (12%) much more efficiently than glucose (0.010.03%). Under these labeling conditions, conversion of [4-13 C]mannose to other monosaccharides found in N-glycans or to glycogen is undetectable (Table 1). In order to assess the contribution of mannose (to glucose) in glycogen and glycolysis, control fibroblasts were labeled with either 5 mM [1,2-13 C]Glc/50 µM Man (shown as 13 C-Glc) or 5 mM Glc/1 mM [1,2-13 C]Man (shown as 13 C-Man) for 24 h, respectively. Under the first condition, 13 C-Man is undetectable in glycogen or in glycolytic products. At a 20-fold higher concentration, mannose is clearly metabolized identically to glucose since it shows a similar distribution in glycogen, and in pyruvate, lactate and alanine in the medium ( Fig.  4A and B).
For assessing the mannose contribution to other sugars in N-glycans, control fibroblasts were labeled with 5 mM [1,2-13 C]Glc and various concentration of [4-13 C]Man, ranging from 50 to 1,000 µM for 24 h. Sugars (mannose, galactose, and N-acetylglucosamine) in N-glycans were analyzed by GC/MS as their corresponding aldonitrile derivatives. At 50 µM, mannose contribution to Man of N-glycans (Man M) was about 2530%; however, increasing mannose to 1 mM, shows that it can completely replace glucose-derived Man (Man G ) in N-glycans (Fig.  4C). Mannose can also contribute to galactose and N-acetylglucosamine in N-glycans. It is undetectable at 50 µM; however, at 1mM 30% of galactose and 50% of N-acetylglucosamine was derived from Man ( Fig. 4D and E). Mannose presumably competes for glucose transport at 1 mM since it reduces uptake of 2-deoxy-[2-3 H] glucose by 40% (data not shown).
Mannose and Glucose take "metabolic detours" within the cell: applying D 2 O as a tracerScores of previous studies using [2-3 H]mannose indicated two mutually exclusive fates: conversion to Fru-6-P (and loss of 3 HOH) and catabolism via glycolysis or, alternatively, conversion to [2-3 H]Man-1-P and incorporation into glycans. Stable isotope labeling suggests the route is more complex. The MPI epimerase reaction requires loss of an H atom at C-2 of the substrate followed by its replacement with an H atom from water, or D in the presence of D 2 O. To test the assumption that conversion to Fru-6-P is irreversible, we labeled fibroblast cells with [1,[2][3][4][5][6][7][8][9][10][11][12][13] C]mannose in the presence of D 2 O. Unexpectedly, this introduced deuterium (D) was incorporated into about one third of the mannose-derived molecules found in N-glycans (Fig. 5B). This shows that a substantial amount of [1,2-13 C]Man-6-P is not directly incorporated into the glycosylation pathway, but is first converted to [1,2-13 C]Fru-6-P and then reconverted to [1,2-13 C-2-D]Man-6-P (Man F ), before incorporation into N-glycans. (Fig.  5A). This reaction is totally dependent on MPI since the D-labeled product (Man F ) is not produced in Mpi-null mouse embryonic fibroblasts (Fig. 5C). This suggests that previous determinations of the proportion of [2-3 H]mannose incorporated into glycans are probably substantially underestimated, since loss of 3 HOH from Man-6-P was assumed to indicate irreversible commitment to further catabolism.
Multiple Metabolic Pools of Fru-6-PTo investigate the hypothesis of Fru-6-P as metabolic 6 checkpoint for glycosylation, cells were labeled with either [UL-13 C]glucose or [UL-13 C]mannose for 6 h, harvested, extracted and dried. Dried sample were derivatized as oxime-TMS and subjected to GC-MS analysis. Man-6-P, Glc-6-P and Fru-6-P are well separated on GC as their corresponding derivatives (Fig. 6A). The origin of sugar phosphate was determined by characteristic fragments of m/z 459 (for Fru-6-P) or m/z 471 (for Man-6-P/Glc-6-P) and their enrichment from [UL- 13 C]sugar (+3 or +4 mass unit labeling, respectively) (Fig. 6B). As expected, contribution from glucose and mannose to Man-6-P (Man G -6-P, Man M -6-P) was proportional to Man G and Man M found in N-glycans (Fig. 7). However, 13 Cmannose was not detected in Fru-6-P even though the D 2 O experiment showed that a substantial amount of Man-6-P is converted to Fru-6-P prior to N-glycosylation (Fig. 5B). Increasing the concentration of mannose to 500 M still did not show detectable contribution to the total cellular Fru-6-P pool (Fig. 7). We conclude that the Fru-6-P derived from exogenous mannose does not equilibrate with the total cellular pool of Fru-6-P, which is overwhelmingly derived from glucose.

Other
Sources of Mannose for N-glycansPrevious studies suggest that glycogen can serve as a reserve source of mannose for glycosylation (18,19). To test this concept, we pre-incubated control fibroblasts with [1,2-13 C]glucose for 2 days to label >80% of glycogen. These cells were then incubated for 24 h in D 2 O and various concentrations of unlabeled glucose. Mannose derived directly from [1,2-13 C]glucose (Man G ) in N-glycans will increase by +2 mass units, but mannose in glycans derived from glycogen (Man GL ) during the D 2 O chase will have an additional mass unit (D) at the C-2 position since all of it must transit through Fru-6-P via MPI (Scheme 2). However, no detectable (<5%) glycogen-derived label (Man GL ) was seen under standard culture conditions (5 mM glucose) or when glucose was reduced to either 1.0 mM or 0.5 mM (Fig. 8A). Under these conditions of limiting glucose, all cellular glycogen was depleted during the 24-h D 2 O chase. As ER-stress is known to increase Man-6-P from glycogen, we tested an ER-stress inducer DTT; however, it did not increase incorporation of glycogen-derived mannose into N-glycans (Fig. 8A). Cells cultured in the complete absence of glucose and mannose for 12 h had very little metabolic activity, and only under these severe conditions, a small amount of glycogen-derived mannose was seen in N-glycans (Fig. 8A). These results make it unlikely that glycogen routinely provides mannose for glycan synthesis, even when glucose is limiting. Similar experiments using two different hepatoma cell lines, HepG2 and Hep3B, led to similar conclusions (data not shown). We also considered the possibility that gluconeogenesis provides mannose for N-glycans, but found no evidence that any label (<5%) was derived from either labeled pyruvate or glycerol under our standard labeling conditions (Fig. 8B). Increasing the concentration of labeled pyruvave or glycerol to 2mM and extending labeling time to 48h also did not show any contribution of gluconeogesis-derived mannose to N-glycans either (data not shown).

Re-utilization of salvaged mannosePrevious
studies showed that mannose released from Nglycan processing is rapidly transported from the ER/Golgi back into the medium (9,20). However, the fate of mannose derived from glycoprotein degradation is unknown. To address this question, we labeled fibroblasts for 72 h with 0.5 mM [1,2-13 C]mannose to replace approximately 60% of the mannose in glycoproteins with Man M . Following a brief washout in unlabeled medium, cells were incubated for 12 h in D 2 O with 12 Cmonosaccharides. This is sufficient time to turnover ~25% of cellular N-glycans (Fig. 3). Based on the results presented above, reincorporation of mannose into newly synthesized glycoproteins would require Man-6-P and ~50% of those molecules would be converted to Fru-6-P and acquire an additional D at C-2. Analysis of N-glycans showed that no mannose (Man S ) with (D) was found (<5%) under physiological condition (5 mM glucose, 50 µM mannose). In the complete absence of glucose and mannose during the incubation with D 2 O, we did detect a small contribution of mannose from salvaged N-glycans (Man S ). However, under normal conditions, we conclude that the mannose salvaged from N-glycans (Man S ) does not make a significant contribution to newly synthesized Nglycans (Fig. 8C).

7
Direct Mannose contribution in various cell linesCancer cell lines are known to possess extremely active sugar metabolism. To evaluate whether they have a unique mannose flux, a wide variety of tumor cell lines were labeled with [4][5][6][7][8][9][10][11][12][13] C]mannose and [1,2-13 C]glucose to determine the contribution from each monosaccharide (Fig.  9A). Mannose contribution varied from ~10% to ~45% at 50 µM mannose and reached up to 75% at 200 µM mannose. Direct mannose contribution generally correlates with the PMM2: MPI ratio, i.e., more mannose is delivered to glycosylation when the ratio is high (Fig. 9B). However, some exceptions suggest that other factors, such as a putative mannose-preferential transporter, may also influence the contribution of Man M .

DISCUSSION
Small amounts of dietary mannose rescue glycosylation-deficient cells (2), mice (3) and humans (4, 5), suggesting its efficient use for glycosylation. Radio-label tracer studies using [2][3] H]mannose indicate the great majority (98%) is catabolized and only a small portion appears in Nglycans. The low glycan labeling efficiency seems at odds with the remarkable physiological benefits. Which perspective is accurate? [2-3 H]Mannose labeling has shortcomings because it does not indicate: 1) the quantity of mannose incorporated; 2) the relative contributions of glucose and mannose to Man-6-P; or 3) the fate of exogenous mannose.
To address these issues we developed stable isotope-labeled methods using glucose and mannose at physiological concentrations, 5 mM and 50 µM, respectively. We detected the origin of mannose of N-glycans based on mass fragment analysis using GC-MS. This efficient, non-toxic and inexpensive method provides both the actual and relative amounts derived from each monosaccharide.
We showed that the amount of mannose directly incorporated into N-glycans is approximately 50% greater than previously estimated. While conversion of [2-3 H]Man-6-P to Fru-6-P through a cis-enediol intermediate indeed releases 3  Stable isotopes also allowed us to determine unambiguously whether mannose in N-glycans is derived from other sources. We found no evidence that gluconeogenic precursors (pyruvate or glycerol) contribute to N-glycan mannose under physiological conditions. Previous reports suggested that ER stress (21) or reduced exogenous glucose (22) could recruit glycogen for glycosylation, but we could not find any evidence for this. The one exception was when cells were completely deprived of all exogenous glucose and mannose for 12 h. We also showed mannose salvaged from N-glycan degradation did not make a significant contribution to mannose in N-glycans under normal conditions. Stable isotopes can be used to track mannose under non-physiological conditions as well.
Increasing mannose concentration to 1mM shows it is metabolized identically to glucose, with similar contributions to glycogen, and to pyruvate, lactate and alanine in the medium. In the complete absence of glucose, mannose salvaged from Nglycans or glycogen-derived glucose shows small contributions to N-glycans.
The direct contribution of mannose to N-glycans varies among different cell lines. The ratio of PMM:MPI activities is an important determining factor; however other undefined factors may also be important, since some cell lines show a large increases in direct mannose incorporation into glycans when exogenous mannose is increased (Fig. 9A, C2C12, U87, Hep3B,) while others (293T, CaCo2) show only modest increases.
A much higher proportion of transported mannose is used for N-glycosylation (~1.8%) compared to glucose (0.026%) and a lingering question is how a 100-fold lower concentration of exogenous mannose vs. glucose, can contribute nearly 50% of the mannose to N-glycans. One explanation is that mannose employs a high affinity, mannosepreferential or -specific hexose transporter that directs it to glycosylation. Studies using [2-8 localized SGLT-type mannose-specific transporter was described (24) that may reabsorb mannose.
Another possibility is that distinct metabolic pools of Fru-6-P exist for glycosylation, glycogenesis, and glycolysis. Our observation that Man M -6-P and Glc-6-P are both transiently converted to Fru-6-P before appearing in glycoproteins and glycogen, respectively, is puzzling since Fru-6-P would seem to be an "unnecessary" intermediate. It is unclear whether this detour process has physiological significance or results from an incidental mixing of Glc-6-P/Man-6-P metabolizing enzymes. To investigate the hypothesis that different pools of Fru-6-P exist for glycosylation, sugar phosphates were analyzed with stable isotopes to determine their origin. The ratio of Man M -6-P and Man G -6-P in the Man-6-P pool was proportional to Man M and Man G found in N-glycans. However, mannose contribution to Fru-6-P at physiological concentration (50 M) nor at a 10-fold higher concentration (500 M) was not detected. This suggests the presence of a separate pool of Fru-6-P that does not equilibrate with the total cellular pool.
Having separate pools does not necessarily mean they reside in different intracellular locations. For example separate sugar phosphate pools could be generated by the anomeric selectivity of Glc-6-P/Man-6-P metabolizing enzymes. For Glc-6-P, PGM1 (Glc-6-P  Glc-1-P) highly prefers the -anomer leading to glycogen synthesis, but the competing enzyme, PGI, prefers the -anomer to supply Fru-6-P for glycolysis. The ratio of  and -Glc-6-P and rate of inter conversion can influence the metabolic fate (25). For Man-6-P, PMM2 is -anomer specific to form Man-1-P for glycosylation, while the competing enzyme, MPI, uses -Man-6-P to form Fru-6-P for glycolysis. Moreover, -Man-6-P is a weak inhibitor of MPI (26). Thus, the much higher efficiency of mannose use in glycosylation might simply result from α and β-Man-6-P and the anomeric preferences of MPI and PMM2.
In summary, the significant findings of this study employing stable isotopes combined with D 2 O were: 1) the direct contribution of mannose to glycoprotein synthesis has been underestimated by at least 1.5 fold; 2) metabolic origin of mannose into N-glycosylation and fate of mannose is unambiguously identified; 3) PMM2:MPI ratio plays a dominant, but not exclusive role in determining the contribution of mannose to Nglycans.  which is an isotopomer without stable isotope and defined as m0. If mannose is originated from glucose, two additional mass units will be observed (m2) since glucose contains two 13 C at C-1 and 2. The % contribution from glucose was calculated as m2/(m0 + m2). In a similar manner, the % contribution from mannose was calculated as m1/(m0 + m1). Natural abundance heavy isotope was corrected using matrix based probabilistic methods    13 C-Man was adjusted to twice the % incorporation into these metabolites, since one glucose/mannose unit produces two pyruvate/lactate/alanine molecules       Table 1 Mannose and glucose uptake and incorporation into N-glycans and glycogen a Rate of sugar uptake was calculated as described under Experimental Procedures. Man M and Man G were calculated by proportions of 13