N-acetylglucosaminyltransferases and nucleotide sugar transporters form multi-enzyme–multi-transporter assemblies in golgi membranes in vivo
Branching and processing of N-glycans in the medial-Golgi rely both on the transport of the donor UDP-N-acetylglucosamine (UDP-GlcNAc) to the Golgi lumen by the SLC35A3 nucleotide sugar transporter (NST) as well as on the addition of the GlcNAc residue to terminal mannoses in nascent N-glycans by several linkage-specific N-acetyl-glucosaminyltransferases (MGAT1-MGAT5). Previous data indicate that the MGATs and NSTs both form higher order assemblies in the Golgi membranes. Here, we investigate their specific and mutual interactions using high-throughput FRET- and BiFC-based interaction screens. We show that MGAT1, MGAT2, MGAT3, MGAT4B (but not MGAT5) and Golgi alpha-mannosidase IIX (MAN2A2) form several distinct molecular assemblies with each other and that the MAN2A2 acts as a central hub for the interactions. Similar assemblies were also detected between the NSTs SLC35A2, SLC35A3, and SLC35A4. Using in vivo BiFC-based FRET interaction screens, we also identified novel ternary complexes between the MGATs themselves or between the MGATs and the NSTs. These findings suggest that the MGATs and the NSTs self-assemble into multi-enzyme/multi-transporter complexes in the Golgi membranes in vivo to facilitate efficient synthesis of complex N-glycans.
KeywordsGolgi apparatus Enzyme complexes Membrane organization Glycosylation
Glycosylation is the most common modification of proteins and lipids, and crucial for multicellular life given its roles in a plethora of basic cellular functions. For example, glycosylation helps proteins to fold correctly, ensures their optimal transport to the cell surface, and mediates a variety of interactions between cells, cells and the extracellular matrix and certain pathogens that use glycans as specific recognition and attachment sites. Therefore, glycosylation is under strict control to allow faithful synthesis of thousands of structurally unique and functionally important glycans on glycoproteins and glycolipids.
Processing of high mannose N-glycans to complex N-glycans commences in the cis/medial-Golgi. It relies on the activity of several distinct N-acetylglucosaminyltransferases termed MGAT1, MGAT2, MGAT3, MGAT4A,4B, MGAT5A,5B and possibly MGAT6  that each add N-acetylglucosamine (GlcNAc) from UDP-GlcNAc to a specific mannose in the core of a N-glycan. In brief, MGAT1 adds the first GlcNAc residue to the C-2 of the α1-3Man in the core of Man5GlcNAc2. Once this step takes place, α-mannosidase II (MAN2A1 or MAN2A2) removes the terminal α1-3Man and α1-6Man residues to form a GlcNAcMan3GlcNAc2N-glycan. Consequently, MGAT2 adds a GlcNAc residue to the C-2 of the α1-6Man, paving the way for additional branching to generate tri- and tetra-antennary N-glycans at C-4 of the core α1-3Man and at C-6 of the core α1-6Man by MGAT4A (or MGAT4B) and MGAT5A (or MGAT5B), respectively.
The substrate required in these processes is UDP-N-acetylglucosamine (UDP-GlcNAc). Given the synthesis of UDP-GlcNAc and other nucleotide sugars in the cytosol  (or in the nucleus in the case of the CMP-sialic acid ), they need to be delivered across the Golgi/ER membrane by relevant nucleotide sugar transporters (NSTs) . These transporters are multi-spanning transmembrane proteins that act as antiporters by exchanging the nucleotide sugar with the corresponding nucleotide monophosphate. The latter is produced following the transfer of a monosaccharide by the relevant glycosyltransferase (GTase)  and the decomposition of the relevant NDP to NMP and inorganic phosphate by nucleotide diphosphatases (NDPases) .
Although GTases and NSTs are often presented as independent players in glycosylation, recent evidence indicates that the enzymes and transporters are functionally organized into highly specific and dynamic complexes both in yeast, plants and mammalian cells [5, 7, 8, 15, 26, 30, 34, 37, 38]. Complex formation for the GTases involves both homomeric and heteromeric interactions, the latter occurring selectively between sequentially acting enzymes within each glycosylation pathway. The formation of enzyme heteromers appears to be strictly dependent on, and regulated by, Golgi micro-environmental factors such as pH (ion) homeostasis . Enzyme heteromers appear also to be functionally more important than the homomers, as their formation increases their enzymatic activity [9, 10, 11, 15]. Enzyme interactions are also mediated mainly by direct contacts between their catalytic domains , a phenomenon that is anticipated to increase both the efficiency and accuracy of glycan biosynthesis, e.g. via prevention of intervention by competing enzymes . Such GTase heteromers have been detected in all major glycosylation pathways and in the different Golgi sub-compartments . Of these, the medial Golgi GTases have been shown to form mega-dalton size assemblies , yet specific interactions have been demonstrated separately only between MGAT1, MGAT2 and an alpha-mannosidase II [24, 25].
NSTs similarly form homomeric complexes (either dimers or higher-order homo-oligomers) in the Golgi compartment. For example, homomeric transporter assemblies have been detected in the case of a UDP-N-acetylgalactosamine transporter , a GDP-fucose transporter , a GDP-mannose transporter [1, 14, 29], a UDP-galactose transporter (SLC35A2)  and UDP-N-acetylglucosamine transporter (SLC35A3) . Structural data on a GDP-mannose transporter also support its tendency to homo-oligomerize . NSTs have also been shown to form heteromeric complexes with each other. For example, the UDP-GlcNAc and UDP-Gal transporters were found to interact in vivo . Very recently, it was shown that MGAT1, MGAT2, MGAT4B and MGAT5 can also separately interact with both a UDP-GlcNAc transporter and a UDP-Gal transporter [18, 20].
To understand better how the GTases and NSTs are functionally organized in the Golgi membranes, we addressed here in detail mutual interactions between the medial-Golgi MGATs, or the NSTs, or those between the MGATs and the NSTs. In the study, we utilized FRET, BiFC and BiFC-based FRET high-throughput interaction screens that allow parallel analyses of multiple protein–protein interactions including ternary ones in thousands of cells in a single experiment. Our data reveal several novel and specific MGAT or NST assemblies in the Golgi membranes of live cells. For the first time, we were also able to identify ternary complexes between the MGATs or between the MGATs and the NSTs. Our findings thus provide direct evidence for the existence of multi-enzyme/multi-transporter assemblies in the Golgi membranes that not only may suffice for Golgi retention, but also to facilitate efficient branching and processing of high-mannose N-glycans to complex N-glycans.
Materials and methods
Construction of expression plasmids
Cell maintenance and transfections
COS-7 cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum, 4 mM l-glutamine, 100 U of penicillin/ml and 100 µg of streptomycin/ml. COS-7 cells were transiently transfected with expression plasmid(s) using Fugene6™ (Promega) according to the manufacturer’s instruction as described elsewhere [10, 11]. For BiFC analysis, cells were subjected to the immunofluorescence staining after 18 h of the transfections. For FRET and BiFC-based FRET analyses, cells were seeded onto 96-well plates (Ultra Perkin Elmer). 18 h post-transfection. After additional 5 h, cells were fixed for 15 min with 4% paraformaldehyde in PBS buffer, washed with PBS before the FRET measurements.
FRET interaction assays
The FRET assay system is an energy transfer system dependent on the distance between one fluorescent donor and one fluorescent acceptor molecule. In each case, the two fluorophores are selected based on their emission and excitation spectra so that the donor’s emission spectrum overlaps with the acceptor excitation spectrum. When the donor molecule is excited, and if the donor (emission) is sufficiently close to the acceptor, an energy transfer will take place, resulting in generation of the emission signal from the acceptor, indicating a positive interaction. At the same time, the emission from the donor fluorophore decreases due to energy transfer to an acceptor fluorophore. If the donor is sufficiently far away from the acceptor (> 10 nm), then energy transfer cannot take place and no signal will be emitted, indicating a negative interaction.
Bimolecular fluorescence complementation (BiFC) and indirect immunofluorescence
Cells were transfected with relevant BiFC enzyme constructs tagged N-terminally with non-fluorescent mVenus fragments (VN and VC), and C-terminally with the HA epitope tag or the FLAG tag and subjected to immunofluorescence staining 18 h post-transfection. To confirm the expression and Golgi localization of the different non-fluorescent VN and VC fusion constructs, we applied the previously described staining protocol [16, 33]. N-terminal VN fusion proteins were detected using the C-terminal HA-epitope tag antibody (monoclonal mouse anti-HA, 1:500 dilution, Sigma-Aldrich), whereas the VC fusion proteins were detected using the rabbit anti-FLAG antibodies (1:250 dilution, Sigma-Aldrich). As secondary antibodies, goat anti-mouse conjugated with Alexa Fluor 405 and goat anti-rabbit conjugated with Alexa Fluor 594 (1:500 dilution; Molecular Probes) were employed. After staining, cells were imaged using the Zeiss LSM700 confocal unit attached to Axio Observer.Z1 microscope, 63 × PlanApo oil immersion objective, and appropriate filter sets for Cerulean, YFP and Alexa Fluor 594 fluorophores. The BiFC signals were quantified from single and double transfected cells using the Operetta High-Content Imaging System. Only cells that expressed either the VN and VC (identified by anti-HA- and anti-FLAG antibodies, respectively) or both VN and VC fusion proteins were used for the quantification.
BiFC-based FRET approach
To enable detection of three separate partners simultaneously, we utilized the BiFC-based FRET approach. In brief, cells were transfected with relevant VN, VC and mCherry- or mCerulean-enzyme encoding plasmid constructs, seeded onto the Cell Carrier Ultra™ 96-well plates 18 h post-transfection, and allowed to attach for 5 h before fixation. Data acquisition and quantification were performed with Operetta High-Content Imaging System similarly to that described above.
Statistical significance of the interaction data sets (with at least three replicates and 10–20 × 103 cells/experiment) were evaluated using either the one-way analysis of variance (ANOVA) (p ≥ 0.05 ns; *p < 0.05; **p < 0.01; ***p < 0.001) or the one tailed t test (p ≥ 0.05 ns; *p < 0.05; **p < 0.01; ***p < 0.001). All statistical analyses were performed using Microsoft Excel software.
MGATs interact with each other in a highly specific manner
Nucleotide-sugar transporter interactions
To confirm the interactions, we utilized also the BiFC (bimolecular fluorescence complementation) approach. This method relies on the formation of a fluorescent mVenus protein from its non-fluorescent N- and C-terminal fragments (VN and VC) if brought close enough by interacting partners . Therefore, we co-transfected COS-7 cells with the relevant plasmids and measured the mVenus-BiFC signal in the Golgi region of single- (used as a negative control) and double-transfected cells. Selection of the Golgi regions of interest (ROI) was done after immunofluorescence staining of the fragments with epitope-tag-specific antibodies (see Fig. 3b). We found that all the tested combinations (A4/A4; A2/A4, A3/A4) gave a strong BiFC signal in the Golgi region of double-transfected (both VN and VC) cells, but not in single transfected cells (either VN or VC alone). Quantitation of the signal intensities with the Operetta High-Content Imaging System showed that the BiFC signal intensity of the A4 homodimers and heterodimers with A2 or A3 were significantly higher compared to cells expressing only the N-terminal fragment of the mVenus protein (Fig. 3c). Thus, both our FRET (Fig. 3a) and BiFC (Fig. 3b, c) data show that the NSTs, similarly to MGATs, form heteromeric assemblies in Golgi membranes in vivo.
Direct detection of ternary complexes using a BiFC-based FRET approach
Previous work on medial-Golgi MGATs had established heteromeric interactions separately only between MGAT1 and MGAT2 or MGAT1 and MGAT2 with MAN2A2 in the ER [24, 25]. Maszczak-Seneczko et al. [18, 19, 20] had also shown interactions between the SLC35A3 and SLC35A2 transporters  or between the MGAT1, MGAT2, MGAT4B or MGAT5A and the transporters [18, 20]. Our findings show for the first time multiple and simultaneous interactions between the MGATs themselves and between the MGATs and NSTs. For example, our in vivo interaction and inhibition studies in the Golgi apparatus revealed several distinct MGAT complexes (Fig. 2e), including the ones between MGAT1, MGAT2 and MAN2A2, between MGAT1, MGAT4B and MAN2A2, or between MGAT2, MGAT3 and MAN2A2. Based on these data (Figs. 1 and 2), we suggest that MAN2A2 serves as a central hub for the MGAT interactions, as it can bind, in a competitive manner, to three different MGAT sub-complexes (MGAT1–MGAT2, MGAT2–MGAT3 and MGAT4B–MGAT1). Therefore, MAN2A2 seems either to have only one binding interface for each of these three MGAT sub-complexes. Alternatively, it may have separate but interdependent binding interfaces for each of them, in which the binding of one sub-complex modulates the other interface so that another sub-complex cannot simultaneously bind to MAN2A2. In either case, however, the MGATs were able to form ternary complexes with MAN2A2. This was also confirmed using the BiFC–FRET approach (Fig. 4b). The specificity of the interactions was confirmed in each case using negative controls, indicating that the interactions are not driven by overexpression of the constructs. In further support, some of the interactions including the one between MGAT1 and MGAT2 have been detected earlier by other means [24, 25].
From the functional point of view, our current data, however, do not allow to discern how these sub-complex assemblies function in N-glycan processing. If the assemblies are static, the acceptor N-glycan is then expected to shuttle between the different sub-complexes to undergo stepwise processing, in which case each sub-complex should recognize the acceptor separately. This may be mediated by MAN2A2 as it is present in all sub-complexes detected in our MGAT interaction screens. Therefore, it may act as a central hub being able to recruit other enzymes and the acceptor substrate that needs further processing. In this case, it is curious that we did not identify any interaction partners for MGAT5, suggesting that it must recognize its acceptor by itself. An alternative possibility also is that the various sub-complexes (i.e. MGAT1, MGAT4B and MAN2A2 or MGAT2, MGAT3 and MAN2A2) may process distinct N-glycan precursors that differ in the number of the antennae added to the high mannose N-glycan. In this scenario, the function of the MGAT1/MGAT2/MAN2A2 complex would thus be to synthesize only biantennary N-glycans, while the others would be involved in making tri- or tetra-antennary N-glycans. A third possibility exists and may involve dynamic exchange of sub-complex constituents. In this case, the acceptor may stay bound to the primary complex (MAN2A2–MGAT1–MGAT2), whose composition then gradually changes by replacement of complex constituents with other enzymes. For example, MGAT1 in the primary complex could be replaced by MGAT3 (or MGAT3–MGAT2 sub-complex) that then adds the bisecting N-acetylglucosamine to an already partly processed N-glycan. Likewise, MGAT4B (or MGAT4B–MGAT1 complex) that adds GlcNAc at C-4 of the core α1-3Man to yield another tri-antennary N-glycan could replace MGAT2 (or MGAT2–MGAT1 complex, respectively) in the primary ternary complex. How such replacements between the complex constituents are coordinated during the processing steps remain unclear but may depend on each GlcNAc residue added to an N-glycan.
We also noticed that MGAT1/2 interact not only with the respective transporter, SLC35A3, but also with SLC35A2 and SLC35A4. It is currently unclear why MGATs interact with the latter two, given that the former is a galactose transporter and the latter is an unknown transporter with no substrate sugar identified yet. Yet, one possibility that may explain these interactions is that the sugar specificities of the transporters may not be absolute  and might be regulated by the enzymes attached to them. Alternatively, it is important to realize (particularly in the case of the A2 transporter) that the interactions are mediated by their transmembrane and/or cytosolic domains (in contrast to enzyme interactions that are mediated mainly by their catalytic domains). This suggests that this interaction is not functionally relevant for the catalytic activity of MGAT1 or MGAT2. Rather, we think that this interaction may relate to acceptor “channeling” to the next enzyme (B4GALT1), a phenomenon that may require later enzyme (B4GALT1) recycling along with its A2 transporter to an earlier compartment, as the Golgi maturation model assumes. Thereby, A2/B4GALT1 complex can become close to MGAT’s TM and cytosolic domains, helping B4GALT1 to retrieve the acceptor from the MGAT1’s catalytic domain for further processing. After giving up the acceptor, MGAT1/2 would be able to recycle into the cis-cisternae and turn it to a new medial Golgi compartment. In our experimental system, such delivery of the acceptor likely does not happen due to the lack of acceptor molecules relative to overexpressed enzyme present, whereby the A2/MGAT1/MGAT2 complexes keep accumulating. Further work is needed to test whether this scenario is correct or not.
To summarize, the data described in this report provide direct evidence for the existence of several multi-enzyme/multi-transporter complexes in Golgi membranes. Such assemblies likely result from the tendency of the GTases and the NSTs to self-assemble with each other, explaining the co-existence of different NSTs with the same nucleotide sugar specificity, as each one of them could deliver the nucleotide sugar only to a distinct enzyme or set of enzymes. Such assemblies are also functionally important as they seem to facilitate more efficient synthesis of complex type N-glycans in the medial-Golgi . In addition, the ability of the GTases and the NSTs to self-assemble into higher order oligomers may also serve not only glycosylation but also Golgi residency, Golgi membrane organization and Golgi stack morphology. It is anticipated that such assemblies also include other Golgi resident proteins including a phosphodiesterase and perhaps proteins associated with Golgi membrane trafficking. These data thus form the basis for future studies aimed at defining all the players in these Golgi membrane enzyme–transporter assemblies and understanding how their formation and functions are regulated at the level of Golgi membranes in living cells.
We wish to thank the National Science Centre (NCN), Krakow, Poland for the funding (grant no. 2014/15/N/NZ1/00492). We also thank the Academy of Finland for funding to SK (grant no. 285232, 2015). PS research visit in Oulu was supported by grant no. 2016/20/T/NZ3/00534 from the National Science Centre (NCN), Krakow, Poland. The authors also greatly acknowledge prof. Pamela Stanley (New York, USA) for many helpful comments on the manuscript. Open access funding provided by University of Oulu including Oulu University Hospital.
- 3.Capasso JM, Hirschberg CB (1984) Mechanisms of glycosylation and sulfation in the Golgi apparatus: evidence for nucleotide sugar/nucleoside monophosphate and nucleotide sulfate/nucleoside monophosphate antiports in the Golgi apparatus membrane. Proc Natl Acad Sci USA 81:7051–7055CrossRefGoogle Scholar
- 4.Coates SW, Gurney T, Sommers LW, Yeh M, Hirschberg CB (1980) Subcellular localization of sugar nucleotide synthetases. J Biol Chem 255:9225–9229Google Scholar
- 12.Hernandez FP and Sandri-Goldin RM (2010) Head-to-tail intramolecular interaction of herpes simplex virus type 1 regulatory protein ICP27 is important for its interaction with cellular mRNA export receptor TAP/NXF1. MBio. https://doi.org/10.1128/mbio.00268-10
- 18.Maszczak-Seneczko D, Sosicka P, Kaczmarek B, Majkowski M, Luzarowski M, Olczak T, Olczak M (2015) UDP-galactose (SLC35A2) and UDP-N-acetylglucosamine (SLC35A3) transporters form glycosylation-related complexes with mannoside acetylglucosaminyltransferases (Mgats). J Biol Chem 290:15475–15486. https://doi.org/10.1074/jbc.M115.636670 CrossRefGoogle Scholar
- 21.Mo S, Chiang S, Lai T, Cheng Y, Chung C, Kuo SCH, Reece KM, Chen Y, Chang N, Wadzinski BE, Chiang C (2014) Visualization of subunit interactions and ternary complexes of protein phosphatase 2A in mammalian cells. PLoS One 9:e116074. https://doi.org/10.1371/journal.pone.0116074 CrossRefGoogle Scholar
- 34.Schachter H (1986) Biosynthetic controls that determine the branching and microhetero-geneity of protein-bound oligosaccharides. Adv Exp Med Biol 64:163–181Google Scholar
- 37.Varki A, Cummings RD, Esko JD, Stanley P, Hart GW, Aebi M, Darvill AG, Kinoshita T, Packer NH, Prestegard JH, Schnaar RL, Seeberger PH (eds) (2017) Essentials of glycobiology, 3rd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NYGoogle Scholar
OpenAccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.