Insights into fucose metabolism: SLC35C1-independent fucosylation discriminates against mannose-derived GDP-fucose

Mutations in the SLC35C1 gene, encoding the Golgi GDP-fucose transporter, cause leukocyte adhesion deciency II (LADII). Fucosylation improvement in LADII patients treated with fucose suggests the existence of an SLC35C1-independent route of GDP-fucose transport, which still remains a mystery. Here, we developed and characterized a human cell-based model decient in the SLC35C1 activity. The knockout cells displayed low but detectable levels of fucosylation. Strikingly, the fucosylation defect was almost completely reversed upon treatment with millimolar concentrations of fucose. Even if fucose was supplemented at nanomolar concentrations, it was still incorporated into glycans by the knockout cells. We also found that the SLC35C1-independent transport preferred the salvage pathway over the de novo pathway as a source of GDP-fucose. Our results imply that the Golgi systems of GDP-fucose transport discriminate between the substrate pools obtained from different metabolic pathways, which suggests a functional connection between nucleotide sugar transporters and nucleotide sugar synthetases. the wild-type and knockout cells, (v) investigation of N-glycan fucosylation in response to various ranges of exogenous fucose concentrations, (vi) single-point and time-course analyses of intracellular GDP-fucose concentration in control and fucose-fed cells, (vii) determination of the contribution of de novo and salvage GDP-fucose biosynthetic pathways to fucose incorporation into N-glycans in the wild-type and knockout cells and (viii) assessment of a potential cellular genetic response to fucose supplementation. observed These results prove the role of SLC35C1 in the process of mucin-type O-glycan fucosylation and goes in line with the observations made for the N-glycans. Importantly, supplementation with 5 mM fucose for 24 h restored fucosylation of O-glycans in the single SLC35C1 and double SLC35C1/SLC35C2 knockout cells. As a follow-up on these remarkable increases in the GDP-fucose concentration in response to fucose supplementation, a measurement of the intracellular concentration of fucose was undertaken. Using an enzymatic assay we have determined the intracellular concentration of fucose at a level of ~1 mM in the SLC35C1-knockouts supplemented with 5 mM fucose. From these numbers one can postulate an approximate equilibrium of 5 mM / ~1 mM / ~0.2 mM for the extracellular fucose / intracellular fucose / intracellular GDP-fucose. GDP-fucose transport Golgi transport biosynthetic this total volume), containing released, free fucose, were combined and transferred to scintillation vials. Glycans bound to the column were released from spin columns with 3 elutions, each with 150 µL of 60% actetonitrile/water/0.5% TFA. Fractions were combined and transferred to scintilation vials. The last step conrmed the 100% rate of fucosidase digestion with no detectable radioactivity in released N-glycans.


Introduction
Fucose is an abundant component of many N-and O-glycans as well as some glycolipids. Except for O-fucosylation, where it is the rst sugar in a sequence, fucose is always a terminal sugar in the structure of oligosaccharides. Fucose can be attached to other sugars or proteins via one of four types of glycosidic bonds: α-1,2, α-1,3, α-1,4 and α-1,6. In N-glycans, fucose is predominantly α-1,6linked to the rst (Asn-bound) N-acetylglucosamine residue. Such type of fucose is referred to as a core fucose and it is incorporated into the oligosaccharide structure only after the attachment of at least one N-acetylglucosamine residue to mannose. For a broader overview of the consecutive steps of N-glycan processing the reader is kindly referred to [1].
Fucosylated oligosaccharides have many biologically relevant functions. AB0 blood group antigens are among the best known fucosylated glycans. O-linked fucose may affect certain ligand-receptor interactions involved in signal transduction. Fucose decorates certain oligosaccharides that are exposed on the surface of leukocytes and serve as ligands for selectins. Binding of the latter to the former initiates the leukocyte adhesion cascade, a multi-step process of leukocyte migration to the site of injury or infection.
Incorporation of fucose into glycoconjugates is mediated by fucosyltransferases, which are predominantly Golgi-resident type II membrane proteins. Fucosyltransferases use the active form of fucose, i.e. GDP-fucose, as a substrate. GDP-fucose is synthesized in the cytoplasm by the two different pathways (shown in Fig. 1). The main source of this nucleotide sugar is the so-called de novo pathway which is estimated to provide 90-95% of the total GDP-fucose pool in the cell [2,3]. The primary substrate for this three-step pathway is mannose. First, this monosugar is converted into GDP-mannose, then GDP-D-mannose 4,6-dehydratase (GMDS) converts GDP-mannose to GDP-4-keto-6-deoxymannose. This keto intermediate is then converted into GDP-fucose by an epimerase/reductase enzyme complex termed the FX protein or GDP-L-fucose synthase (also known as GFUS, FCL, SDR4E1 or TSTA3) [4].
The second, alternative way of GDP-fucose synthesis is the so-called salvage pathway. This route uses cytoplasmic pool of free fucose, which on the one hand is recovered from glycoconjugates by lysosomal α-fucosidase activity, and on the other hand is supplied from the environment. Little is known about the fucose transport system from the extracellular space to the interior of the cell, but it appears to function through facilitated diffusion through speci c channel(s) [5]. The rst step of the salvage pathway involves an ATP-dependent synthesis of fucose-1-phosphate by the fucokinase (FUK; also known as FCSK) enzyme. GDP-fucose pyrophosphorylase (FPGT), the second enzyme acting in the pathway, catalyzes the conversion of fucose-1-phosphate and GTP to GDP-fucose [6]. As already mentioned, it is estimated that only 5-10% of the total GDP-fucose pool is synthesized by this pathway [2,3].
To reach lumenally oriented catalytic centers of fucosyltransferases, GDP-fucose has to be transported across the Golgi membrane. This function is thought to be played by GDP-fucose-speci c nucleotide sugar transporters (NSTs), which are hydrophobic type III membrane proteins with 6-10 transmembrane domains. NSTs are believed to act as antiporters; an imported nucleotide sugar is exchanged with the corresponding nucleoside monophosphate [7]. After reaching the Golgi lumen, the activated monosugar constitutes a substrate for a respective glycosyltransferase, which attaches the monosugar to an acceptor, whereas the released nucleoside diphosphate is broken down into a nucleoside monophosphate and an inorganic phosphate [8].
In mammals, the role of the main GDP-fucose transporter is attributed to the product of the SLC35C1 gene, which was identi ed by two independent groups in 2001 [9,10]. Although SLC35C1 is thought to be the major transporter in mammals, GDP-fucose has also been shown to translocate through its homolog, SLC35C2. SLC35C2 was shown to localize predominantly in the Golgi, but a small subset of it was found in the endoplasmic reticulum (ER) and ERGIC (ER-Golgi intermediate compartment). SLC35C2 is speci cally required for O-fucosylation of certain proteins including the Notch receptor, which in turn does not require the SLC35C1 activity [11,12].
The concept of using exogenous fucose to improve fucosylation in LADII patients was rst raised in 1998. Pioneering experiments were performed with lymphoblastic cells obtained from one of the patients bearing the Arg147Cys mutation in the SLC35C1 amino acid sequence. Supplementation of the cell culture with 10 mM fucose for 5 days triggered an increase in the number of fucosylated structures [24]. These results paved the way to the concept of using fucose supplementation in patients themselves. Since 1999, several LADII patients were treated with oral fucose. For some of them such treatment was successful, resulting in an improvement in psychomotor development and a reduction in neutrophil count [22,25]. However, in the case of two LADII patients fucose supplementation failed to improve their condition [20,26]. The administered doses of fucose varied from 25 to 2000 mg/kg body weight [20,25,26]. An improvement in fucosylation was also achieved when the LADII patient-derived broblasts were cultured in the presence of 0.1-10 mM fucose [18].
To explain the effectiveness of the fucose treatment in responsive LADII patients, it was postulated that the mutant SLC35C1 variants display some residual transport activity [9]. Therefore, a therapy causing an increase in the cytosolic concentration of this nucleotide sugar could potentially allow to overcome the insu cient transporting activity of the defective SLC35C1 variants. However, a direct proof for an increase in the cytosolic GDP-fucose concentration in the fucose-fed cells bearing mutations in the SLC35C1 gene has not been presented to date. An alternative hypothesis to explain the e ciency of the oral fucose therapy could be the existence of an accessory GDP-fucose transport system, however, no such route has been identi ed so far.
In 2007, a knockout-based study conducted in mice revealed that the fucose treatment works even in the absence of SLC35C1 [27]. These ndings for the rst time suggested the existence of an SLC35C1-independent GDP-fucose transport into the Golgi, as the observed effect could no longer be explained by a residual activity of the SLC35C1 mutants. It was hypothesized that the SLC35C1 de ciency could be overcome by the SLC35C2 activity, but this has never been proved. In general, no studies that could possibly explain this phenomenon in bigger detail have been undertaken so far.
Since the responsiveness of the cells completely devoid of the SLC35C1 activity to fucose treatment has not been elucidated to date, here we have undertaken a more-in-depth investigation of this intriguing effect. Within our general interest in the metabolism of fucose and protein fucosylation, speci c goals included: (i) generation of single SLC35C1 and double SLC35C1/SLC35C2 knockouts, (ii) optimization of fucose feeding conditions, (iii) establishment of a method allowing quanti cation of the extent of N-glycan fucosylation, (iv) structural characterization of N-and O-glycans produced by the wild-type and knockout cells, (v) investigation of Nglycan fucosylation in response to various ranges of exogenous fucose concentrations, (vi) single-point and time-course analyses of intracellular GDP-fucose concentration in control and fucose-fed cells, (vii) determination of the contribution of de novo and salvage GDP-fucose biosynthetic pathways to fucose incorporation into N-glycans in the wild-type and knockout cells and (viii) assessment of a potential cellular genetic response to fucose supplementation.
In this study we have developed an approach coupling exoglycosidase digestion with HPLC to quantify the percentage of the corefucosylated N-glycans with a great precision. We showed that in the knockout cells the core fucosylation of N-glycans is greatly reduced but can be nearly completely restored by supplementing the cells with millimolar concentrations of fucose. Supplementation with 5 mM fucose caused a signi cant increase in the intracellular GDP-fucose concentration, regardless of the presence/absence of GDP-fucose transporters. Next, by feeding low (nanomolar) concentrations of radioactive fucose and mannose into the culture medium we were able to follow the metabolic fate of the supplemented compounds and attribute radioactivity to fucosylated structures using our HPLC-based method. Surprisingly, the SLC35C1 knockouts showed a dramatic impairment of the ability to incorporate mannose-derived radioactivity into the fucosylated N-glycans. In sharp contrast, they were able to incorporate even higher quantities of radiolabeled fucose than the wild-type cells. To explain these observations we hypothesize that in the SLC35C1 knockouts the salvage pathway is preferred over the de novo pathway as a source of GDP-fucose for fucosylation of N-glycans. We have also investigated potential genetic response of the SLC35C1 knockouts to fucose treatment. No major gene up/downregulations were observed. Finally, we excluded SLC35C2 as an alternative supplier of GDP-fucose for N-glycan fucosylation.

Results
Generation of the single SLC35C1 and double SLC35C1/SLC35C2 knockouts To develop a research model for our study, an SLC35C1 knockout (C1KO) was generated in two human cell lines, i.e. HEK293T and HepG2 using the CRISPR/Cas9 approach. Additionally, a double SLC35C1/SLC35C2 knockout (C1/C2KO) was developed in HEK293T cells to evaluate the in uence of SLC35C2 protein on the N-and O-glycosylation in the SLC35C1 knockout cells and to examine the effect of fucose treatment in cells lacking both GDP-fucose transporters. The knockouts were con rmed by RT-PCR performed on total RNA and PCR performed on genomic DNA ( Fig. S1A and S1B). The SLC35C1 knockout was additionally veri ed on the protein level using an anti-SLC35C1 antibody by western blotting (Fig. S1C) as well as by double staining with the latter antibody and fucosespeci c Aleuria aurantia lectin (AAL, Fig. S1D).
To show that the knockout phenotype was speci cally associated only with the disruption of the SLC35C1 gene, we have also developed SLC35C1-de cient clones stably over-expressing a recombinant, HA-tagged SLC35C1 (in both HEK293T and HepG2 cell lines). The overexpression was con rmed by western blotting performed on cell lysates using an anti-HA antibody (Fig. S1E). The obtained stable transfectants were also subjected to immunostaining to con rm the Golgi localization of the over-expressed protein (Fig. S1F).

Optimization of fucose supplementation conditions
In order to establish a system to study the effect of fucose supplementation on the restoration of fucosylation in the SLC35C1 knockouts, feeding parameters, i.e. fucose concentration in the culture medium as well as duration of the treatment, had to be optimized. In the rst stage we employed a dot blot analysis using AAL lectin ( Fig. 2A, 2B). Several concentrations of fucose were tested, i.e. 0.5, 1, 5 and 10 mM (concentration range inspired by [18]). We found that after 5 days, 5 mM concentration was su cient to substantially restore fucosylation of glycoconjugates in the SLC35C1 knockouts. AAL stainings performed on the SLC35C1 knockouts cultured in the presence of 5 mM fucose for 0-48 h demonstrated that the level of fucosylation gradually increased over time and an optimum (i.e. fucosylation compared to the wild-type) was reached after 24 h. After another 12 h, however, the cells became overloaded with aggregated fucose-containing structures (Fig. 2C).
Since the lectin-based analysis does not allow for quanti cation of the observed fucosylation defects/rescue effects, we have established a new, robust method, based on exoglycosidase digestions followed by HPLC separation, that allows quanti cation of the degree of the α-1,6-fucosylation (core fucose). In our approach, we used speci c glycosidases to reduce all multi-branched complex-type N-glycan species into a simple biantennary (GlcNAc) 2 (Man) 3 structure, either fucosylated or non-fucosylated (Fig. 2D,   2E). Then by taking a ratio of the area of the fucosylated peak over the sum of areas of both peaks (fucosylated and nonfucosylated) we have determined the percentage of fucosylated structures for each sample. To con rm that the fucosylated peak contains only (GlcNAc) 2 (Man) 3 Fuc structure, the α-fucosidase digestion was performed (Fig. 2F) By employing our method we tested supplementation of selected fucose concentrations (50 µM -100 mM) for different time periods (1-5 days) in the wild-type and SLC35C1 knockout HEK293T cells. Feeding the cells with 50 µM or 1 mM fucose even for as long as 5 days was not su cient to restore the wild-type α-1,6-fucosylation level in the SLC35C1 knockout cells. Only supplementation of the SLC35C1 knockout cells with 5 mM fucose restored fucosylation phenotype to a level comparable to the wild-type, but at the same time it did not affect core fucosylation in the wild-type cells. At the same time, supplementation with a very high fucose concentration (100 mM) did not increase the level of core fucosylation in the wild-type cells (Fig. 2G). Therefore, for the nal experiments, the conditions of 5 mM fucose and 24 h were selected.
Supplementation of fucose in the knockout cells signi cantly increases α-1,6-fucosylation of N-glycans We quanti ed α-1,6-fucosylation of N-glycans in all modi ed cell lines. From the analysis of the HEK293T cells we could conclude that: (i) the native level of core fucosylation in the non-fed wild-type cells oscillates around ~75% and is compromised to ~8-10% in the single SLC35C1 and double SLC35C1/SLC35C2 knockouts, (ii) ectopic expression of the HA-tagged SLC35C1 in the single SLC35C1 knockouts restores the wild-type phenotype, (iii) nearly complete α-1,6 fucosylation (~60%) can be achieved in both single SLC35C1 and double SLC35C1/SLC35C2 knockouts upon administration of 5 mM fucose for 24 h (Fig. 3A). As anticipated, similar observations were made for the HepG2 cells (Fig. 3B). MALDI-TOF analysis of N-glycan structures con rmed those results, showing the presence of a fucosylated structure in the SLC35C1 knockouts and restoration of fucosylated structures after fucose feeding (Fig.  S2).
During the separation and puri cation of N-glycans, there is a possibility of contamination with the oligosaccharides derived from serum glycoproteins contained in the culture medium. To address this problem, the N-glycans derived from an engineered, ectopically expressed and secreted reporter glycoprotein, Secreted Alkaline Phosphatase (SEAP), were also analyzed. The method for SEAP isolation e ciently prevents contamination with serum glycoproteins [28]. From the analysis of α-1,6-fucosylated N-glycans derived from the SEAP overexpressed in the wild-type and knockout HEK293T cells we concluded that in the knockouts α-1,6-fucosylation of SEAP-derived N-glycans is still present (~3-5%) and supplementation of 5 mM fucose for 24 h restore α-1,6-fucosylation in these cells. Interestingly, the level of α-1,6-fucosylation observed in the wild-type and knockout cells was higher than it was in the wild-type before supplementation, which could be a speci c response of the overexpressed SEAP on the supplementation (Fig. 3C). As anticipated, similar observations were made for the HepG2 cells (Fig. 3D).

Supplementation of fucose restores fucosylation of O-glycans in the knockout cells
Certain mucin-type O-glycans are also fucosylated. The effect of GDP-fucose transport de ciency on this type of oligosaccharides was only investigated using lectins [27]. Therefore, we have additionally analyzed cellular mucin-type O-glycans using the CORA strategy [29].

Supplementation of fucose increases intracellular GDP-fucose concentration
Although it is anticipated that in the SLC35C1-de cient cells the exogenous fucose e ciently enters the cells and becomes converted into GDP-fucose, to our best knowledge this phenomenon was neither quanti ed. For separation and quantitative analysis of nucleotide sugars we employed a modi ed ion-pair solid-phase extraction and HPLC strategy developed by Räbinä et al. [30]. Our results demonstrated that in the non-fed HEK293T cells the intracellular concentration of GDP-fucose oscillated around ~5 µM, regardless of the genotype (wild-type or single/double knockout). This shows that GDP-fucose transporter de ciency of the SLC35C1 knockout did not cause a GDP-fucose accumulation. This is in contrast to the research conducted on the SLC35A1 (CMP-sialic acid transporter) knockout, for which such an effect was observed [31]. Importantly, upon fucose supplementation, GDP-fucose concentration increased ~40-50-fold in all HEK293T cell lines, up to ~200-250 µM (Fig. 5A). Similar experiment was performed for HepG2 cells where the basal concentrations of GDP-fucose were slightly higher (~10 µM) and the effect of supplementation was somewhat less pronounced (~15-fold, up to ~150 µM) (Fig. 5B). Based on these ndings we concluded that the externally supplemented fucose e ciently enters the cytoplasm and is being converted to GDP-fucose in all tested cell lines.

Supplementation of fucose causes dynamic changes in fucose metabolism
As a follow-up on these remarkable increases in the GDP-fucose concentration in response to fucose supplementation, a measurement of the intracellular concentration of fucose was undertaken. Using an enzymatic assay we have determined the intracellular concentration of fucose at a level of ~1 mM in the SLC35C1-knockouts supplemented with 5 mM fucose. From these numbers one can postulate an approximate equilibrium of 5 mM / ~1 mM / ~0.2 mM for the extracellular fucose / intracellular fucose / intracellular GDP-fucose.
Subsequently, in order to study the GDP-fucose turnover, the time-course evaluation of the intracellular GDP-fucose concentration was performed in the HEK293T wild-type and SLC35C1 knockout cells over the course of 24 h. As shown in Fig. 5C, GDP-fucose concentrations close to the maximum levels of ~250 µM were reached as soon as after 6-12 h since the beginning of the supplementation. However, when the supplementation was terminated, GDP-fucose concentration remained stable for ~8 h and then started to gradually decline towards basal levels (Fig. 5D). These observations were then correlated to the glycophenotype investigated using AAL uorescent staining (Fig. 5E). After 24 h of supplementation fucosylation was strongly pronounced which was manifested by an e cient lectin binding (Fig. 5E, 0h). Withdrawal of the monosaccharide from the culture medium caused a gradual reduction of fucosylation which was no longer detectable after 48 h.
Having discovered that supplementation of 5 mM fucose for 24 h led to a drastic increase in the intracellular GDP-fucose concentration, we next examined how different fucose concentrations will affect the intracellular GDP-fucose concentration and core fucosylation of N-glycans. Therefore, we cultured the wild-type and SLC35C1 knockout HEK293T cells in the presence of 0 -5 mM fucose for 24 h and determined the GDP-fucose concentrations for each of the tested fucose concentrations. A visible increase in the GDP-fucose concentration required the supplementation of at least 0.2 -1 mM fucose (Fig. 5F). Then we have demonstrated that in the SLC35C1 knockouts core fucosylation of N-glycans increases proportionally to the concentration of the intracellular GDP-fucose ( Fig. 5G). The nding that increasing GDP-fucose concentration in the wild-type cells does not have any effect on the core fucosylation of N-glycan goes in line with our previous results. Our data demonstrate that a signi cant increase in the intracellular GDP-fucose concentration is necessary to restore core fucosylation of N-glycans in the SLC35C1 knockouts.
Our data showed that GDP-fucose derived from the salvage pathway is e ciently utilized in the SLC35C1 and SLC35C1/C2 knockout cells. By supplementation of cells with small concentration of radiolabeled [6-3 H]fucose, we were able to track the biosynthesis of GDP-fucose via the salvage pathway and incorporation of fucose into N-glycans. We showed that the radioisotope incorporated into core fucosylated N-glycans produced by all tested cell lines (Fig. 6A, B, C). It means that even small amounts of exogenous fucose can be incorporated into core fucosylated N-glycans in the SLC35C1 and SLC35C1/C2 knockouts. Based on the known total radioactivity of the fucose substrate fed to the cells and on the measured radioactivity in the isolated fraction of the core fucosylated N-glycans, one can estimate the e ciency of the radioisotope incorporation ( Fig. 6D, E). The fact that the higher percentage of core fucose was incorporated into N-glycans by the knockout cells may suggest a greater contribution of the GDP-fucose derived from the salvage pathway in the knockout cells as compared to the wild-types. We also observed that the biosynthesis of GDP-fucose in the salvage pathway occurs in both the wild-type and knockout cells (Fig. 6F, G, H). Altogether, these results may suggest the preferential usage of the salvage pathway-produced GDP-fucose for the N-glycan fucosylation in the SLC35C1 knockout cells.

Supplementation of millimolar mannose does not increase the intracellular GDP-fucose concentration
We discovered that only when we increase intracellular GDP-fucose, a restoration of the core fucosylation of N-glycans in the SLC35C1 knockout cells is observed. We next tested if 5 mM mannose (another source of GDP-fucose in the cells) feeding for 48 h will increase the intracellular GDP-fucose concentration. In both the wild-type and knockout cells mannose supplementation caused an increase in the intracellular GDP-mannose concentration (from ~100 µM to ~220 µM in the wild-type and from ~180 µM to ~330 µM in the knockout; Fig. 7A). In contrast, no mannose to GDP-fucose conversion was observed in the mannose-fed cells, regardless of their genotype (Fig. 7B). Having found that feeding the cells with 5 mM mannose does not cause any increase in the intracellular GDP-fucose concentration we next decided to examine whether radiolabeled mannose supplemented in low (nanomolar) concentrations will be converted into GDP-fucose, i.e. to verify, whether the de novo pathway works properly in the analyzed cells. Importantly, both the wild-type and SLC35C1 knockout cells e ciently converted the supplemented radiolabeled mannose into GDPfucose (Fig. 7C, D).
The de novo and salvage pathways differentially contribute into N-glycan fucosylation in the SLC35C1 knockout cells To compare the relative contribution of different GDP-fucose biosynthesis pathways into the N-glycan fucosylation, the wild-type and SLC35C1-de cient HEK293T cells were cultured in the presence of nanomolar concentrations of either radioactive fucose (salvage pathway) or radioactive mannose (de novo pathway). However, apart from conversion into GDP-fucose, mannose can be also processed into other nucleotide sugars, e.g. GDP-mannose, and as such incorporated into the complex type N-glycans. To address this problem, we have established a new approach in which the N-glycans are rst digested with α-fucosidase, then the post-reaction mixture is passed through a graphite column and nally the radioactivity of the ow-through is measured. This strategy allows us to speci cally assign the detected radioactivity to fucose present in N-glycans.
Our results showed a decrease in the incorporation of fucose derived from the GDP-fucose synthesized via the salvage pathway into N-glycans in the SLC35C1 knockout cells, but still the extent of fucose incorporation was more than 23% of the wild-type (Fig. 8A). Surprisingly, incorporation of fucose derived from the GDP-fucose synthesized via the de novo pathway (i.e. from mannose) turned out to be nearly completely abolished (6% of the wild-type) (Fig. 8B).

qPCR analysis of expression levels of selected glycosylation-related genes
There is a possibility that some fucosylation-related genes become up-or down-regulated by the exogenously added fucose. Therefore, to complement our metabolic studies, expression levels of selected glycosylation-related genes were analyzed in the control and fucose-fed SLC35C1-de cient cell lines. For the analysis of a larger subset of genes, a ULP probe-based assay was designed ( Fig S4). Here, 91 different genes, including NSTs, glycosyltransferases and biosynthetic enzymes involved in metabolism of different monosaccharides (galactose, N-acetylglucosamine, mannose, xylose and fucose) were investigated. No statistically signi cant differences could be assigned for any of the tested genes.
To analyze differences in gene expression between the wild-type and SLC35C1 knockouts in HEK293T and HepG2 cells a SYBR Green-based qPCR assay was performed for six selected genes, i.e. FCL, GMDS, FUK, FPGT1, FPGT2 and SLC35C2 (Sup tab.1, 2). Similarly to the previous ndings, no statistically signi cant differences between the wild-type and SLC35C1 knockout HEK293T and HepG2 cells (both control and supplemented with exogenous fucose) were observed.

Discussion
In this study we have con rmed the existence of SLC35C1-independent GDP-fucose transport routes in the Golgi complex and determined the intracellular fate of the exogenously supplied fucose. Moreover, we have shown that the SLC35C1-dependent route mainly transports GDP-fucose provided by the de novo biosynthetic pathway, whereas the SLC35C1-independent route nearly exclusively translocates GDP-fucose derived from the salvage pathway.
The SLC35C1-de cient mammalian model was for the rst time developed and characterized by Hellbusch and coworkers in 2007 [27]. In that study, treatment of cells from different mouse organs with exogenously supplied fucose partially restored glycoprotein fucosylation, which encouraged the authors to propose the existence of an alternative, SLC35C1-independent GDP-fucose transport mechanism. However, the improvement in fucosylation was demonstrated using only a single method, i.e. lectin-based ow cytometry, and no intermediate metabolic steps, e.g. conversion of fucose into GDP-fucose, have been investigated.
The study presented here in multiple aspects extends the previously conducted research. First of all, the improvement in fucosylation caused by supplementation of the culture medium with millimolar concentrations of fucose was precisely determined to be ~77% by our self-developed quantitative HPLC-based approach. Moreover, we showed that also nanomolar concentrations of exogenous fucose are su cient to observe its incorporation into N-glycans in the knockout cells. Furthermore, we determined the extent to which exogenous fucose is transported into the cytoplasm and how it contributes to the increase of the concentration of the intracellular GDP-fucose (the extracellular fucose, intracellular fucose and intracellular GDP-fucose were shown to exist in a 5 mM / 1 mM / 0.2 mM equilibrium). Finally, we showed that the SLC35C1-independent Golgi GDP-fucose transport systems preferentially use the nucleotide sugar pool derived from the salvage pathway, which we consider the main outcome of this study.
In this study we have developed a quantitative HPLC-based approach allowing for the very precise determination of the level of core fucosylation of N-glycans. This form of fucosylation, mediated by a single Fut8 enzyme, is present in many cell types including HEK293T and HepG2 which were selected as our model. Moreover, it is easy to study due to the well-established procedures of Nglycan isolation and HPLC analysis. As shown by our data, the level of core-fucosylated N-glycans in the wild-type HEK293T and HepG2 cell lines is high (~80%) but does not reach 100%, giving a possibility to potentially observe a further increase in fucosylation, i.e. over-fucosylation, upon fucose treatment. On the other hand, in these cell lines the level of core-fucosylated structures is high enough to make a clear distinction between the wild-type and the SLC35C1 knockout phenotypes. To conclude, the choice of cell lines was dictated by the high percentage of core-fucosylated N-glycans they produce, which in turn was demanded by the quantitative HPLC-based approached we employed.
Using our quantitative HPLC-based approach we showed that the core N-glycan fucosylation is signi cantly compromised in the SLC35C1-de cient cells. However, some residual fucosylation could still be detected in these cell lines. The SLC35C2 protein is also speci c for GDP-fucose. Thus, its activity could potentially compensate for the lack of SLC35C1 in the single knockout cells. Here we showed that this is not the case as knocking out the SLC35C2 gene in the SLC35C1-de cient cells did not make the fucosylation defect any more severe. There results suggest that there must be yet another system responsible for the SLC35C1-independent GDPfucose import into the Golgi lumen.
Strikingly, supplementation of the culture medium with millimolar concentrations of fucose caused a substantial improvement of core N-glycan fucosylation in both single SLC35C1 and double SLC35C1/SLC35C2 knockout cells. In contrast, in the wild-type cells fucose treatment did not cause any further improvement in fucosylation, which is in line with the results obtained by Moriwaki et al. [33] who did not observe any increase in AAL reactivity with glycoproteins produced by the wild-type Hep3B cells supplemented with up to 5 mM fucose. This nding additionally supports our assumption that SLC35C1 mainly utilizes GDP-fucose produced by the de novo pathway, whereas GDP-fucose synthesized via the salvage pathway does not appear to be an optimal source of this nucleotide sugar in the wild-type cells, even if the latter is present in a large excess. We also hypothesize that the residual fucosylation detectable in the knockout cells is derived from GDP-fucose produced by the salvage pathway, which is not the main source of this nucleotide sugar in the wild-type cells. We believe that if the SLC35C1-independent GDP-fucose transport system used the nucleotide sugar substrate produced by the main biosynthetic pathway (i.e. de novo), the level of core fucosylated N-glycans in the knockout cells would be much higher.
It should be emphasized that fucose treatment of the LADII patients and SLC35C1-de cient cells can only be successful if the following requirements are ful lled: (i) fucose must e ciently enter the cells, (ii) after entering the cell fucose must be readily converted into GDP-fucose via the salvage pathway and (iii) GDP-fucose must reach the Golgi lumen to become available for the catalytic centers of fucosyltransferases. Importantly, quanti tation of intracellular GDP-fucose concentration in fucose-fed SLC35C1 knockout cells was not attempted in the previous studies. To the best of our knowledge, the changes in the GDP-fucose content in response to exogenously supplied fucose was examined only by Moriwaki et al. in the wild-type Hep3B cells [32] and, more recently, by Sosicka et al. in the wild-type HepG2, CHO and Huh7 cells [33]. In the rst study, the range of analyzed fucose concentrations was between 0 and 5000 µM. However, these results cannot be compared with the ones obtained in this study as both procedures differed in several aspects including the cell line used and the method for nucleotide sugar detection (Moriwaki et al. [33] used an indirect, enzyme-based detection and did not attempt to determine intracellular GDP-fucose concentration, whereas we performed a direct detection followed by estimation of intracellular GDP-fucose concentration). On the other hand, Sosicka et al. tested only one concentration of exogenous fucose (50 µM) [33]. Importantly, neither of these studies attempted to determine the intracellular GDPfucose concentration in fucose-fed SLC35C1 knockout cells, which is crucial for our understanding of the responsiveness of these cells to exogenous fucose.
Here, we determined GDP-fucose concentrations in the control and fucose-fed wild-type and knockout cells. We found that the baseline concentrations of GDP-fucose were similar in the wild-type and knockout cells, whereas a remarkable increase could be observed upon fucose treatment in all the cell lines analyzed. In native conditions, the salvage pathway was shown to play only negligible role in the overall GDP-fucose production. Here we show that the capacity of this pathway is very high, i.e., the corresponding enzymes are able to process signi cantly greater amounts of the primary substrate than are normally present in the cell. However, our data exclude the possibility that the exogenously added fucose triggers a strong up-regulation of expression of the enzymes acting in this pathway as we did not observe any signi cant changes in the relative levels of the corresponding transcripts between the control and fucose-fed SLC35C1 knockout cells.
Our results clearly show that the therapeutic approaches for CDGs associated with impaired NST activity should aim for a drastic increase in the intracellular concentration of the corresponding nucleotide sugar substrates. It should be noted that treatment of these CDGs by oral monosaccharide supplementation can only be successful if the Km of an enzyme initiating the corresponding salvage pathway is su ciently high. Otherwise, the treatment would simply be ineffective. Fucokinase is the rst enzyme acting in the salvage pathway of the GDP-fucose biosynthesis. Importantly, in the case of fucokinase puri ed from pig kidney Km for fucose was determined to be as high as 27 µM [32]. This explains why sub-millimolar and millimolar concentrations of exogenous fucose are required to e ciently stimulate GDP-fucose synthesis and rescue N-glycan fucosylation in SLC35C1-de cient cells.
Given the fact that fucose treatment nearly completely rescued the fucosylation defect in the SLC35C1-de cient cells, it appears surprising that fucose-based therapy turned out to be ineffective for some of the LADII patients. There are two possible explanations for this phenomenon. First, the non-responsive LADII patients might bear additional mutations outside the SLC35C1 gene that affect the performance of the fucosylation machinery. The other possible explanation is that the presence of certain mutant SLC35C1 variants is somehow more deleterious to the cells than the absence of the protein.
Regardless of the improvement of fucosylation in the SLC35C1-de cient cells treated with millimolar concentrations of fucose, we have also demonstrated that these cells are able to incorporate fucose supplemented in nanomolar concentrations. This proves that in the knockout cells the salvage pathway is able to fuel the SLC35C1-independent GDP-fucose transport system even at extremely low fucose concentrations.
Since increasing the intracellular concentration of GDP-fucose via fucose treatment improved core fucosylation of N-glycans, we wondered if it would be possible to boost the production of this nucleotide sugar by feeding the cells with millimolar concentrations of mannose given the fact that the de novo pathway of GDP-fucose synthesis utilizes GDP-mannose as a substrate. However, regardless of the cell line analyzed, mannose treatment did not cause any signi cant changes in the GDP-fucose concentration. This can be partially explained by the fact that mannose is converted to a variety of different metabolites. Although in mannose-fed cells a statistically signi cant increase in the GDP-mannose concentration was demonstrated, this did not translate into a similar increase in the GDP-fucose level. We therefore hypothesize that both in the wild-type and knockout cells the de novo pathway operates with a nearly maximum e ciency and it is impossible to improve its performance by increasing the amount of its primary substrate, i.e., GDP-mannose. Based on these ndings it can be concluded that treatment of LADII patients with mannose wouldn't be an attractive alternative to fucose feeding.
Based on our results we propose the existence of three different GDP-fucose transport systems in the Golgi membrane. The rst one is SLC35C1-dependent and mainly utilizes the nucleotide sugar pool derived from the de novo pathway. The other two are not dependent on SLC35C1 and mainly use the nucleotide sugar pool synthesized in the salvage pathway. However, the rst of them works even under very low (nanomolar) concentrations of exogenous fucose that do not cause any increase in the GDP-fucose concentrations, whereas the other requires much higher concentrations of this nucleotide sugar that can only be obtained by feeding the cells with sub-millimolar and millimolar concentrations of fucose. The existence of the latter SLC35C1-independent transport system is strongly supported by the non-linear dependence of the GDP-fucose content on the concentration of exogenous fucose. We hypothesize that this route may not be physiological and is not exclusively speci c for GDP-fucose.
The hypothesis raised above assumes the existence of distinct, independent, separate cytosolic pools of GDP-fucose. Such a phenomenon was recently proposed by Sosicka et al. [33] who elegantly demonstrated that different fucosyltransferases utilize distinct GDP-fucose pools derived from distinct fucose sources. We believe that this effect could be executed by a selective cooperation of the different GDP-fucose transport systems with distinct fucosyltransferases (similar to e.g. SLC35C2, which was shown to speci cally support O-fucosylation [11,12]). But the question arises how different Golgi transport systems could discriminate between the GDP-fucose pools derived from distinct metabolic pathways? This might require a physical proximity (or even association) between the enzymes acting in the individual pathways and the corresponding transport systems. Although the predominant Golgi localization of nucleotide sugar transporters is well established, little is known about the precise subcellular distribution of nucleotide sugar synthases. The study performed by Coates et al. postulated cytoplasmic localization of the GDPmannose and UDP-glucose pyrophosphorylases [35]. This appears to hold true for all the other nucleotide sugar synthases except for the CMP-sialic acid synthetase (CMAS) whose nuclear localization was shown in the same study. However, it is widely believed that cytoplasmic enzymes catalyzing consecutive reactions in the individual pathways are not randomly distributed. Instead, they are rather sequestered to speci c cytoplasmic subcompartments where they form functional assemblies that support substrate channeling [36]. It cannot be excluded that the enzymes acting in the GDP-fucose biosynthetic pathways also display such a tendency.
Importantly, GDP-fucose was shown to inhibit the de novo pathway in humans [37] and bacteria [38]. At the same time, this pathway is thought to be the main source of GDP-fucose in mammalian cells. Therefore, an immediate delivery of the GDP-fucose synthesized by the de novo pathway to the site of its ultimate utilization (i.e., the Golgi lumen) would be highly bene cial as it would ensure the e cient course of the GDP-fucose biosynthesis by shifting the equilibrium of the reaction. Such a scenario could be possible if the enzymes acting in the de novo pathway and the main GDP-fucose transporter, i.e., SLC35C1, were located nearby. Although the mammalian fucokinase was shown to be nearly completely inhibited by 60 µM GDP-fucose [34], we did not observe such phenomenon, as we were able to increase the intracellular concentration of GDP-fucose up to ~200 µM by fucose treatment. This may suggest that the GDP-fucose formed in the salvage pathway does not accumulate in the site of its synthesis but instead is immediately channeled to subcompartments in which it is subsequently utilized. Interestingly, the interactions between a plant UDPglucose 4-epimerase and two UDP-galactose transporters were recently reported [39]. Therefore, it is highly likely that also the enzymes acting in the de novo pathway of GDP-fucose biosynthesis are near to the SLC35C1 protein in mammalian cells.
In this study we have con rmed the existence of the alternative, SLC35C1-independent GDP-fucose transport systems in the Golgi complex. We also have shown that SLC35C2 activity is dispensable for this transport to occur. We demonstrated that the exogenously supplemented fucose e ciently enters the cells and is readily converted to GDP-fucose via the salvage pathway. Strikingly, the SLC35C1-de cient cells were virtually unable to incorporate fucose derived from GDP-fucose produced by the de novo pathway. There are several questions that still need to be answered, e.g. what is the molecular identity of the SLC35C1-independent GDP-fucose transport systems and how different Golgi transport systems discriminate between the nucleotide sugar pools derived from different biosynthetic pathways. Nevertheless, we strongly believe that the ndings obtained in this study initiate a brand new chapter in our perception of fucosylation.

Materials And Methods
Cell culturing and gene inactivation HEK293T cells purchased from ATCC (American Type Culture Collection, CRL-3216) and HepG2 cells purchased from the collection of the Department of Cancer Immunology (Hirszfeld Institute of Immunology and Experimental Therapy, Polish Academy of Sciences) were cultured in Dulbecco's Minimum Eagle Medium (DMEM High Glucose, Sigma-Aldrich) supplemented with 10% fetal bovine serum, 100 U/mL penicillin and 100 µg/mL streptomycin. Cells were kept at 37°C under 5% CO 2 .
Ready-to-use Santa Cruz Biotechnology CRISPR-Cas9 kit was used to inactivate SLC35C1 gene in HEK293T and HepG2 cell lines. Cells were transfected with a mixture of human SLC35C1 double nickase plasmids (sc-410008-NIC-2) using the FuGENE HD transfection reagent (Promega) according to manufacturer's protocol. Next, cells were cultured in DMEM complete medium supplemented with 1 µg/mL of puromycin for three weeks to select transfected cells. After that time, clones were isolated and checked for the presence of the SLC35C1 transcript using one-step reverse transcription PCR (RT-PCR) with the HEK293T and HepG2 wild-type cells as the controls. Analysis of genomic DNA derived from the SLC35C1-de cient cells was also performed. Primers used in the reactions were listed in Table 1. Next step in con rmation of the SLC35C1 gene knockout was western blotting.
Inactivation of the SLC35C2 gene in the SLC35C1 knockout HEK293T cells was performed in the same manner. Santa Cruz Biotechnology CRISPR-Cas9 kit with mixture of human SLC35C2 double nickase plasmids (sc-409264-NIC-2) was applied. Con rmation of the SLC35C2 gene knockout was performed at both RNA and genomic DNA levels with using the HEK293T wild-type cells as a control. Used primers are listed in Table 1. pSelect-zeo-mcs plasmid vector (InvivoGen) using SalI (New England Biolabs) restriction enzyme. Subsequently, HA-SLC35C1 mRNA was modi ed within protospacer adjacent motif region recognized by Cas9 protein by introducing a silent mutation using Q5 Site-Directed Mutagenesis Kit (New England Biolabs) according to manufacturer's protocol. This step was necessary to protect the HA-SLC35C1 mRNA from the activity of the Cas9 protein in the knockout cells. The primers used for mutagenesis were listed in Table 1.
The SLC35C1 knockout HEK293T and HepG2 cells were transfected with the obtained vector using the FuGENE HD transfection reagent (Promega) according to manufacturer's instructions. Cells were then cultured in DMEM complete medium with the addition of zeocin (400 µg/mL for HEK293T and 200 µg/mL for HepG2) until stable clones were isolated. Cells stably expressing the HAtagged SLC35C1 were identi ed by immuno uorescent staining with an anti-HA primary antibody followed by a secondary antibody conjugated with an Alexa Fluor dye. Additionally, con rmation of the expression of the desired protein was done by western blotting.

Western blotting
Cell lysates were separated in 10% SDS-PAGE gels and transferred onto nitrocellulose membranes (Amersham). After the transfer membranes were blocked. Then, speci c fragments of membranes were incubated with appropriate primary and secondary antibodies (listed in Table 2). Steps of blocking, antibodies incubation, washing and signal detection were performed as described previously [40]. Subcellular localization of the endogenous GDP-fucose transporter and HA-tagged recombinant SLC35C1 protein was determined using primary and secondary antibodies listed in Table 3. Cells were immunostained as described previously [41]. In the case of determination of fucosylation level with biotinylated lectin in the wild-type and SLC35C1-de cient cells, procedure was modi ed. After xation, cells were permeabilized for 5 min at RT using 0.1% Triton X-100 in TBS. Non-speci c binding sites were blocked with 3% BSA in TBS for 1 h at RT. Then, biotinylated Aleuria aurantia lectin (AAL; Vector Laboratories cat # B-1395-1, 1:300) diluted in blocking solution containing 1 mM CaCl 2 and 1 mM MnCl 2 was added for 1 h at 37°C. Cells were washed with TBS. Slides were then incubated with streptavidin-Cy3 (Sigma Aldrich, 1:500) solution for 1 h at 37°C. The rest of the protocol remained unchanged. The resulting samples were analyzed using a Leica SP8 confocal microscope and the obtained images were processed using an ImageJ software.

N-glycan analysis
Control and fucose-fed (5 mM, 24 h) cells were lysed and treated with acetone to concentrate proteins. Proteins were then dissolved and enzymatically deglycosylated. Released glycans were puri ed, labeled with 2-AB and analyzed as described previously [40]. Obtained glycans were subjected to MALDI-TOF analysis.

O-glycan analysis
Control and fucose-fed (5 mM, 24 h) cells were cultured in medium with 5% fetal bovine serum with an addition of peracetylated Oglycan precursor (Ac 3 GalNAcBn) for 3 days. Subsequently, O-glycans contained in the culture medium were puri ed according to an adopted method described by Kudelka et al [29]. The obtained glycans were subjected to MALDI-TOF analysis.

MALDI-TOF mass spectroscopy analysis of N-and O-glycans
Prior to MALDI-TOF analysis, 2-AB-labeled N-glycans were desialylated using α2-3,6,8,9-neuraminidase A (New England Biolabs) and puri ed in the manner reported previously [40]. Both N-and O-glycans were analyzed in a positive-ion mode as described previously [42].

Analysis of SEAP-derived N-glycans
Cells were transiently transfected with 6xHis-SEAP construct using the FuGENE HD transfection reagent (Promega) according to manufacturer's recommendations. Medium containing secreted protein was collected. The SEAP reporter glycoprotein was subjected to further puri cation following the protocol described by Olczak and Szulc [28]. Then, isolated N-glycans were labeled with 2-AB and analyzed as described previously [40].

Preparation of RNA
Total RNA was extracted from ~3.2 x 10 6 cells (HEK293T, HepG2) using the Total RNA Mini Kit (A&A Biotechnology, Gdynia, Poland). Isolated RNA was treated with DNase I (A&A Biotechnology) and puri ed using the Clean-Up RNA Concentrator Kit (A&A Biotechnology). RNA integrity was veri ed spectrophotometrically.

SYBR Green-based quantitative RT-PCR
Reverse transcription was carried out on 1 µg of total RNA using a SensiFAST cDNA Synthesis Kit (Bioline). PCR was carried out using the Luna Universal qPCR Master Mix (New England Biolabs) and a LightCycler 96 instrument (Roche). The ampli cation reaction comprised initial denaturation at 95°C for 1 min and followed by 45 ampli cation steps (denaturation at 95 ℃ for 15 s; primer annealing and extension at 60 ℃ for 30 s). The melting curves were analyzed to monitor the quality of PCR products. Relative quanti cation of expression of respective genes was determined by the ΔΔCq method using human glyceraldehyde-3-phosphate dehydrogenase (GAPDH; NM_002046.7) as a reference. Three independent experiments (biological replicates) were performed to test each experimental condition and each sample was run in three technical replicates. Hence, in total 3x3=9 runs were executed per one experimental condition. No template controls were included on each reaction plate to check for contamination. Negative controls consisting of untranscribed RNA (no-RT controls) were performed to check for genomic DNA contamination. All primers used in this study are listed in Table 4. cDNA was produced using Transcriptor First Strand cDNA Synthesis Kit (Hoffmann-La Roche, Basel, Switzerland) according to the manufacturer's instructions. Total RNA (4 µg) extracted from HEK293T cells was mixed with 2 µL of 50 pmol/µL anchoredoligo(dT)18 primer and supplemented with PCR-grade water up to 26 µL nal volume. The template-primer mixture was incubated in a thermal block cycler with a heated lid for 10 minutes at 65 ℃ to denature RNA secondary structures and cooled on ice.
Quantitative PCR procedure qPCR reaction mixture was assembled at room temperature by mixing cDNA (40 µL) obtained in the previous step with PCR-grade water (960 µL) and 2x concentrated LightCycler 480 Probes Master (Roche; 1 mL) and pipetted into a 96-well RealTime ready Custom Panel plate (Roche; 20 µL per well). The plate was sealed with LightCycler 480 Sealing Foil (Roche) and rotated in a horizontal position for 5 minutes at 1000 rpm to dissolve primers and probes lyophilized on the bottom of the Custom Panel plate wells. The plate was spun down in a centrifuge equipped with a 96-well plate swing-bucket rotor to remove any potential air bubbles and subjected to the qPCR experiment using LightCycler 96 instrument (Roche). The PCR reaction was initiated by a 10-minute preincubation at 96 ℃ and followed by 45 ampli cation steps (95 ℃ for 10 s; 60 ℃ for 30 s; 72 ℃ for 1 s).
Custom Panel qPCR plates qPCR assays (primer pairs and probes) were purchased from Roche and supplied in a lyophilized form in 96-well plates. The exact con guration of the plate including assay positions on the plate, assay IDs and gene names, sequences of forward and reverse primers and ULP probe numbers is provided in Table 5. Assays were veri ed by the supplier to ful ll a number of quality criteria including: (i) PCR e ciency 2.0 +/-0.2 (equals 100 +/-10%), (ii) Cq of highest cDNA concentration ≤34, (iii) Linear dynamic range of at least 3 logs, (iv) High ampli cation speci city, no side products in gel analysis, (v) Sigmoidal ampli cation curve, (vi) Fluorescence intensity of ampli cation curves between 5 and 50 uorescence units. The analysis was performed using 4 different algorithms, i.e. GeNorm, NormFinder, BestKeeper and delta-Cq method. The most stable consensus gene was beta-1,4-glucuronyltransferase 1 (B4GAT1).

Quantitative PCR data analysis
To test the null hypothesis that feeding cells with 5 mM fucose does not affect the expression of the tested glycosylation-related genes the SLC35C1-de cient HEK293T cells grown in fucose-supplemented medium (N=4, i.e. 4 biological repetitions or 4 Custom Panel qPCR plates) were compared to the SLC35C1-de cient HEK293T cells supplemented with PBS (also N=4). Relative expression folds were calculated using the ΔΔCq method following Taylor et al. [43]. The results were subjected to the t-test and Benjamini-Hochberg procedure to assess the signi cance of the gene expression differences in the non-fed and fed cells.

Optimization of fucose supplementation conditions
Three supplementation parameters were optimized, i.e. time of supplementation, frequency of medium exchange and concentration of L-fucose. The rst optimizations were carried out in two variants. In the rst variant, the wild-type and SLC35C1 knockout cells were cultured in complete medium (DMEM High Glucose, 10% fetal bovine serum, 100 U/mL penicillin, 100 µg/mL streptomycin) with addition of 0.5 mM, 1 mM, 5 mM and 10 mM L-fucose dissolved in phosphate-buffered saline (PBS) for 5 and 12 days and the medium was exchanged every day. In the second variant, the wild-type and SLC35C1 knockout cells were cultured in complete medium with addition of 0.5 mM, 1 mM, 5 mM and 10 mM L-fucose dissolved in PBS for 5 and 12 days but the medium was replaced every other day. As the control, the wild-type and SLC35C1-de cient cells were kept in complete medium supplemented with PBS. Control cells were also cultured in two variants as the fucose-fed cells. After that, cells were collected, lysed and subjected to dot-blotting with AAL to visualize fucosylation level. Procedure of dot blotting is described in the section below.
In the next step, SLC35C1 knockout cells were cultured on glass 8-well microscope slides (Merck) in complete medium with 5 mM Lfucose for 48 h. Cells were xed after 2, 4, 6, 8, 10, 24, 36 and 48 hour after starting supplementation. Fixed cells were subjected to uorescent staining with AAL lectin according to the procedure mentioned above.
As the last experiment, the wild-type and SLC35C1 knockout cells were treated with different L-fucose concentrations for different times. Cells were collected and lysed. Cell lysates were subjected to N-glycan extraction performed as described above. Puri ed and 2-AB-labeled N-glycans were analyzed for α-1,6-fucosylation as described below.

Dot blotting
Whole cell lysates obtained from untreated and fucose-fed cells were applied directly on nitrocellulose membrane (Amersham). Glycans containing fucose were detected using biotinylated AAL (Vector Laboratories cat # B-1395-1, 1:300) as described previously [40]. Lectin attached to glycans was visualized by streptavidin conjugated to HRP (Vector Laboratories, cat # SA-5014-1, 1:50000) by chemiluminescence reaction using Western Lightning Plus-ECL kit (PerkinElmer). Equal protein loading was demonstrated by staining proteins on nitrocellulose membrane with the Ponceau S solution.
Determination of L-fucose concentration in the fucose-fed SLC35C1-de cient cells Concentration of L-fucose in fucose-fed SLC35C1-de cient HEK293T cells was determined using a commercial kit (L-fucose, Megazyme). Brie y, ~120x10 cells were resuspended in MiliQ water and then sonicated. Next, in order to get rid of proteins from the lysate, ice-cold perchloric acid was added. The mixture was spun down and supernatant was collected. Subsequently, assay sample was prepared by mixing the supernatant with the detection buffer and NADP solution (the latter two were supplied by the kit) in volumes recommended by manufacturer. Mixture was incubated over 3 minutes. After that, L-fucose dehydrogenase suspension (supplied by the kit) was added to the mixture in volume recommended by manufacturer. Finally, the absorbance was measured at 340 nm using a Beckman DU-640 spectrophotometer. Concentration of fucose in the test samples was determined based on a calibration curve prepared for the fucose standard (supplied by the kit).
Radioactive labeling of N-glycans and nucleotide sugars h to analyze N-glycans. Then cells were harvested and subjected to N-glycan preparation followed by analysis of N-glycan α-1,6 fucosylation. For analysis of nucleotide sugars cells were incubated in complete medium with addition of 4 µCi/mL L-[ Analysis of N-glycan α-1,6-fucosylation Puri ed, 2-AB-labeled and dried N-glycan pools were dissolved in 25 µL of certi ed, exoglycosidase free GlycoBuffer 1 (50 mM sodium citrate, pH 6.0; New England Biolabs). Neuraminidase A (40 units per reaction), β-1-4 galactosidase S (16 units per reaction) and β-N-Acetyl-Glucosaminidase S (8 units per reaction) were added, to start digestion. All enzymes were purchased from New England Biolabs. Reactions were performed overnight, at 37°C in air-heated incubator, to avoid evaporation. In these conditions, virtually all complex-type N-glycans were converted to two conserved trimannosyl core structures (fucosylated at the rst Nacetylglucosamine and non-fucosylated). The post-reaction mixtures were separated on GlycoSep N Plus column (Prozyme) at 40°C. High Resolution gradient #1 (Prozyme manual for the column) was applied, and detection was performed with a uorescence detector (330 nm excitation, 420 nm emission) connected to a Perkin Elmer Series 200 HPLC gradient system. The percentage of Nglycan core fucosylation was calculated from peak area of fucosylated trimannosyl core compared to area of peak eluted as nonfucosylated trimannosyl structure, using TotalChrom software (Perkin Elmer).
The same procedure of separation was applied for N-glycans released from cells labeled with 3 H-fucose. In these experiments, fractions of 8 drops (approximately 230 µL) were collected manually and directly transferred to scintillation vials.
Calculations of the extent of radioactive fucose incorporation were based on radioactivity (expressed in cpm) of radioisotope of known concentration and known speci c activity compared to peak areas of standard, 2-AB-labeled, trimannosyl core structures (fucosylated and non-fucosylated) of known concentrations (all standard glycans were purchased from Oxford Glycosystems).

Analysis of N-glycan fucosylation after 3 H-mannose and 3 H-fucose labeling
Puri ed, 2-AB-labeled and dried N-glycan pools released from cells labeled with 3 H--fucose or 3 H-mannose, were resuspended in 25 µL of certi ed, exoglycosidase free GlycoBuffer 1 (50 mM sodium citrate, pH 6.0, New England Biolabs). Then α1-2,4,6 Fucosidase O (New England Biolabs, 4 Units per reaction) was added to start de-fucosylation. The reaction was performed overnight at 37°C. After digestion, the mixtures were complemented with 250 µL of 1% TFA, mixed and applied to the Graphite Spin Columns (Pierce), previously washed two times with 100 µL of 1 M ammonia, primed by 100 µL of acetonitrile and, nally, equilibrated two times with 100 µL of 1% TFA, each step was nished by centrifugation at 2000 g for 1 minute. Applied samples were bound to the graphite resin in closed spin column for 20 minutes, with periodic vortex mixing. After 20-minute incubation, the columns were centrifuged at 2,000g for 3 minutes and washed 3 times with 100 µL of 5% acetonitrile/water/1% TFA solution. The ow-through and washes (approximalely 550 µL of total volume), containing released, free fucose, were combined and transferred to scintillation vials. Glycans bound to the column were released from spin columns with 3 elutions, each with 150 µL of 60% actetonitrile/water/0.5% TFA. Fractions were combined and transferred to scintilation vials. The last step con rmed the 100% rate of fucosidase digestion with no detectable radioactivity in released N-glycans.
Control reactions for the same procedure were also performed, without fucosidase added to GlycoBuffer 1. This con rmed lack of non-enzyme de-fucosylation during overnight incubations.

HPLC separation of nucleotide sugars
Nucleotide sugars were puri ed from collected and frozen cells using protocol previously published [30]. Enriched nucleotide sugar pools were separated in ion-pairing, reverse phase HPLC, as described previously [44] with some changes. The column was changed to Interstil ODS-4, 250 x 3 mm, 3 µm particle size, ow rate was established to 0.3 mL/min. Buffer A (100 mM potassium phosphate, pH 6.4 supplemented with 8 mM tetrabutylammonium hydrogen sulphate) was used for equilibration, Buffer B (70% of A plus 30% acetonitrile) was used as an eluent. Separations were performed using the following gradient: 0% for 20 minutes, 0-72% for 25 minutes, 72%-77% for 10 minutes and, nally, equilibrated at 0% of buffer A for 25 minutes. Nucleotide sugar samples were resuspended in MiliQ water.
The sample volumes of 5 µL or less were injected to start separations. The column was connected to Nexera Shimadzu HPLC system. Detection was performed at 254 nm with SPD_M30A Diode Array Detector equipped with HS (high-sensitivity) quartz cell. Calculations of nucleotide sugar concentrations were performed with LabSolutions Software (Shimadzu). Standard, high purity nucleotide sugars were purchased from Sigma-Aldrich. The absolute quanti cation of nucleotide sugars was performed by comparison of the detected signals to the externally added reference compounds. Next, by referring to the starting number of cells and to the anticipated volume of a single cell (3 and 3.5 pL per cell for HepG2 and HEK293T, respectively), it was possible to estimate the intracellular nucleotide sugar concentrations.
The same procedure of ion-pairing reverse-phase separation was applied for nucleotide sugar pools, previously puri ed from cells labeled with 3 H-fucose and 3 H mannose (as described before). In these experiments, fractions of 5 drops (approximately 150 µL) between 26th and 33rd minute of HPLC gradient, were collected manually and directly transferred to scintillation vials. Retention times under separation conditions described above were approximately 27.2 min. and 31.1 min., for GDP-mannose and GDP-fucose, respectively.

Statistical Analysis
Statistical parameters including data plotted (mean ± SD), P values, and statistical tests used are mentioned in Figure Legends. Statistical analyses were performed using Graphpad Prism 6. Data were analyzed by Student's t-test and Benjamini-Hochberg procedure or Welch's correction and by one-way ANOVA with the Tukey post-hoc test.

Declarations Competing interests
All authors declare no competing interests.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.    Metabolic labeling of the cellular N-glycans with radioactive fucose. (A) Exemplary representative HPLC chromatograms of the digested N-glycans isolated from the indicated cell lines. The uorescence signal from the 2-AB-labeled N-glycans (black solid lines) was overlayed with the radioactivity data (red dots and solid lines). The total radioactivity of the fucosylated N-glycans normalized for the amount of the starting material (total protein) from HEK293T (B) and HepG2 (C) cell lines. Percentage of the radioactive fucose incorporated into the fucosylated N-glycans isolated from HEK293T (D) and HepG2 (E) cell lines. To calculate the values the sum radioactivity measured in the HPLC fractions corresponding to the fucosylated species was divided by the total radioactivity of the fucose added to the culture medium. Comparison of the relative incorporation of the radioactive fucose derived from the salvage pathway in the wild-type (F) and SLC35C1 knockout (G) HEK293T cells. (H) The total radioactivity of the GDP-fucose normalized for total number of HEK293T cells. ns, not signi cant; *, p < 0.05 as determined using two-tailed unpaired t-test with Welch's correction. Data are represented as mean ± SEM. ISTD, internal standard (GDP-glucose).

Figure 7
Quanti cation of the intracellular GDP-mannose and GDP-fucose concentration in cells fed with mannose. Intracellular concentrations of GDP-mannose (A) and GDP-fucose (B) in the wild-type and SLC35C1 knockout HEK293T cells cultured for 48 h in the absence and presence of 5 mM mannose. Comparison of the relative incorporation of the radioactive fucose derived from the de novo pathway in the wild-type (C) and SLC35C1 knockout (D) HEK293T cells. ISTD, internal standard (GDP-glucose). ns, not signi cant; *, p < 0.05; **, p < 0.01 as determined using two-tailed unpaired t-test with Welch's correction. Data are represented as mean ± SEM